Even as concern and skepticism grows over U.S. AI startup OpenAI's buildout strategy and high spending commitments, Chinese open source AI providers are escalating their competition and one has even caught up to OpenAI's flagship, paid proprietary model GPT-5 in key third-party performance benchmarks with a new, free model.

The Chinese AI startup Moonshot AI’s new Kimi K2 Thinking model, released today, has vaulted past both proprietary and open-weight competitors to claim the top position in reasoning, coding, and agentic-tool benchmarks.

Despite being fully open-source, the model now outperforms OpenAI’s GPT-5, Anthropic’s Claude Sonnet 4.5 (Thinking mode), and xAI's Grok-4 on several standard evaluations — an inflection point for the competitiveness of open AI systems.

Developers can access the model via platform.moonshot.ai and kimi.com; weights and code are hosted on Hugging Face. The open release includes APIs for chat, reasoning, and multi-tool workflows.

Users can try out Kimi K2 Thinking directly through its own ChatGPT-like website competitor and on a Hugging Face space as well.

Moonshot AI has formally released Kimi K2 Thinking under a Modified MIT License on Hugging Face.

The license grants full commercial and derivative rights — meaning individual researchers and developers working on behalf of enterprise clients can access it freely and use it in commercial applications — but adds one restriction:

"If the software or any derivative product serves over 100 million monthly active users or generates over $20 million USD per month in revenue, the deployer must prominently display 'Kimi K2' on the product’s user interface."

For most research and enterprise applications, this clause functions as a light-touch attribution requirement while preserving the freedoms of standard MIT licensing.

It makes K2 Thinking one of the most permissively licensed frontier-class models currently available.

Kimi K2 Thinking is a Mixture-of-Experts (MoE) model built around one trillion parameters, of which 32 billion activate per inference.

It combines long-horizon reasoning with structured tool use, executing up to 200–300 sequential tool calls without human intervention.

According to Moonshot’s published test results, K2 Thinking achieved:

• 44.9 % on Humanity’s Last Exam (HLE), a state-of-the-art score;

• 60.2 % on BrowseComp, an agentic web-search and reasoning test;

• 71.3 % on SWE-Bench Verified and 83.1 % on LiveCodeBench v6, key coding evaluations;

• 56.3 % on Seal-0, a benchmark for real-world information retrieval.

Across these tasks, K2 Thinking consistently outperforms GPT-5’s corresponding scores and surpasses the previous open-weight leader MiniMax-M2—released just weeks earlier by Chinese rival MiniMax AI.

GPT-5 and Claude Sonnet 4.5 Thinking remain the leading proprietary “thinking” models.

Yet in the same benchmark suite, K2 Thinking’s agentic reasoning scores exceed both: for instance, on BrowseComp the open model’s 60.2 % decisively leads GPT-5’s 54.9 % and Claude 4.5’s 24.1 %.

K2 Thinking also edges GPT-5 in GPQA Diamond (85.7 % vs 84.5 %) and matches it on mathematical reasoning tasks such as AIME 2025 and HMMT 2025.

Only in certain heavy-mode configurations—where GPT-5 aggregates multiple trajectories—does the proprietary model regain parity.

That Moonshot’s fully open-weight release can meet or exceed GPT-5’s scores marks a turning point. The gap between closed frontier systems and publicly available models has effectively collapsed for high-end reasoning and coding.

When VentureBeat profiled MiniMax-M2 just a week and a half ago, it was hailed as the “new king of open-source LLMs,” achieving top scores among open-weight systems:

• τ²-Bench 77.2

• BrowseComp 44.0

• FinSearchComp-global 65.5

• SWE-Bench Verified 69.4

Those results placed MiniMax-M2 near GPT-5-level capability in agentic tool use. Yet Kimi K2 Thinking now eclipses them by wide margins.

Its BrowseComp result of 60.2 % exceeds M2’s 44.0 %, and its SWE-Bench Verified 71.3 % edges out M2’s 69.4 %. Even on financial-reasoning tasks such as FinSearchComp-T3 (47.4 %), K2 Thinking performs comparably while maintaining superior general-purpose reasoning.

Technically, both models adopt sparse Mixture-of-Experts architectures for compute efficiency, but Moonshot’s network activates more experts and deploys advanced quantization-aware training (INT4 QAT).

This design doubles inference speed relative to standard precision without degrading accuracy—critical for long “thinking-token” sessions reaching 256 k context windows.

K2 Thinking’s defining capability lies in its explicit reasoning trace. The model outputs an auxiliary field, reasoning_content, revealing intermediate logic before each final response. This transparency preserves coherence across long multi-turn tasks and multi-step tool calls.

A reference implementation published by Moonshot demonstrates how the model autonomously conducts a “daily news report” workflow: invoking date and web-search tools, analyzing retrieved content, and composing structured output—all while maintaining internal reasoning state.

This end-to-end autonomy enables the model to plan, search, execute, and synthesize evidence across hundreds of steps, mirroring the emerging class of “agentic AI” systems that operate with minimal supervision.

Despite its trillion-parameter scale, K2 Thinking’s runtime cost remains modest. Moonshot lists usage at:

• $0.15 / 1 M tokens (cache hit)

• $0.60 / 1 M tokens (cache miss)

• $2.50 / 1 M tokens output

These rates are competitive even against MiniMax-M2’s $0.30 input / $1.20 output pricing—and an order of magnitude below GPT-5 ($1.25 input / $10 output).

The rapid succession of M2 and K2 Thinking illustrates how quickly open-source research is catching frontier systems. MiniMax-M2 demonstrated that open models could approach GPT-5-class agentic capability at a fraction of the compute cost. Moonshot has now advanced that frontier further, pushing open weights beyond parity into outright leadership.

Both models rely on sparse activation for efficiency, but K2 Thinking’s higher activation count (32 B vs 10 B active parameters) yields stronger reasoning fidelity across domains. Its test-time scaling—expanding “thinking tokens” and tool-calling turns—provides measurable performance gains without retraining, a feature not yet observed in MiniMax-M2.

Moonshot reports that K2 Thinking supports native INT4 inference and 256 k-token contexts with minimal performance degradation. Its architecture integrates quantization, parallel trajectory aggregation (“heavy mode”), and Mixture-of-Experts routing tuned for reasoning tasks.

In practice, these optimizations allow K2 Thinking to sustain complex planning loops—code compile–test–fix, search–analyze–summarize—over hundreds of tool calls. This capability underpins its superior results on BrowseComp and SWE-Bench, where reasoning continuity is decisive.

The convergence of open and closed models at the high end signals a structural shift in the AI landscape. Enterprises that once relied exclusively on proprietary APIs can now deploy open alternatives matching GPT-5-level reasoning while retaining full control of weights, data, and compliance.

Moonshot’s open publication strategy follows the precedent set by DeepSeek R1, Qwen3, GLM-4.6 and MiniMax-M2 but extends it to full agentic reasoning.

For academic and enterprise developers, K2 Thinking provides both transparency and interoperability—the ability to inspect reasoning traces and fine-tune performance for domain-specific agents.

The arrival of K2 Thinking signals that Moonshot — a young startup founded in 2023 with investment from some of China's biggest apps and tech companies — is here to play in an intensifying competition, and comes amid growing scrutiny of the financial sustainability of AI’s largest players.

Just a day ago, OpenAI CFO Sarah Friar sparked controversy after suggesting at WSJ Tech Live event that the U.S. government might eventually need to provide a “backstop” for the company’s more than $1.4 trillion in compute and data-center commitments — a comment widely interpreted as a call for taxpayer-backed loan guarantees.

Although Friar later clarified that OpenAI was not seeking direct federal support, the episode reignited debate about the scale and concentration of AI capital spending.

With OpenAI, Microsoft, Meta, and Google all racing to secure long-term chip supply, critics warn of an unsustainable investment bubble and “AI arms race” driven more by strategic fear than commercial returns — one that could "blow up" and take down the entire global economy with it if there is hesitation or market uncertainty, as so many trades and valuations have now been made in anticipation of continued hefty AI investment and massive returns.

Against that backdrop, Moonshot AI’s and MiniMax’s open-weight releases put more pressure on U.S. proprietary AI firms and their backers to justify the size of the investments and paths to profitability.

If an enterprise customer can just as easily get comparable or better performance from a free, open source Chinese AI model than they do with paid, proprietary AI solutions like OpenAI's GPT-5, Anthropic's Claude Sonnet 4.5, or Google's Gemini 2.5 Pro — why would they continue paying to access the proprietary models? Already, Silicon Valley stalwarts like Airbnb have raised eyebrows for admitting to heavily using Chinese open source alternatives like Alibaba's Qwen over OpenAI's proprietary offerings.

For investors and enterprises, these developments suggest that high-end AI capability is no longer synonymous with high-end capital expenditure. The most advanced reasoning systems may now come not from companies building gigascale data centers, but from research groups optimizing architectures and quantization for efficiency.

In that sense, K2 Thinking’s benchmark dominance is not just a technical milestone—it’s a strategic one, arriving at a moment when the AI market’s biggest question has shifted from how powerful models can become to who can afford to sustain them.

Within weeks of MiniMax-M2’s ascent, Kimi K2 Thinking has overtaken it—along with GPT-5 and Claude 4.5—across nearly every reasoning and agentic benchmark.

The model demonstrates that open-weight systems can now meet or surpass proprietary frontier models in both capability and efficiency.

For the AI research community, K2 Thinking represents more than another open model: it is evidence that the frontier has become collaborative.

The best-performing reasoning model available today is not a closed commercial product but an open-source system accessible to anyone.

Google Cloud is introducing what it calls its most powerful artificial intelligence infrastructure to date, unveiling a seventh-generation Tensor Processing Unit and expanded Arm-based computing options designed to meet surging demand for AI model deployment — what the company characterizes as a fundamental industry shift from training models to serving them to billions of users.

The announcement, made Thursday, centers on Ironwood, Google's latest custom AI accelerator chip, which will become generally available in the coming weeks. In a striking validation of the technology, Anthropic, the AI safety company behind the Claude family of models, disclosed plans to access up to one million of these TPU chips — a commitment worth tens of billions of dollars and among the largest known AI infrastructure deals to date.

The move underscores an intensifying competition among cloud providers to control the infrastructure layer powering artificial intelligence, even as questions mount about whether the industry can sustain its current pace of capital expenditure. Google's approach — building custom silicon rather than relying solely on Nvidia's dominant GPU chips — amounts to a long-term bet that vertical integration from chip design through software will deliver superior economics and performance.

Google executives framed the announcements around what they call "the age of inference" — a transition point where companies shift resources from training frontier AI models to deploying them in production applications serving millions or billions of requests daily.

"Today's frontier models, including Google's Gemini, Veo, and Imagen and Anthropic's Claude train and serve on Tensor Processing Units," said Amin Vahdat, vice president and general manager of AI and Infrastructure at Google Cloud. "For many organizations, the focus is shifting from training these models to powering useful, responsive interactions with them."

This transition has profound implications for infrastructure requirements. Where training workloads can often tolerate batch processing and longer completion times, inference — the process of actually running a trained model to generate responses — demands consistently low latency, high throughput, and unwavering reliability. A chatbot that takes 30 seconds to respond, or a coding assistant that frequently times out, becomes unusable regardless of the underlying model's capabilities.

Agentic workflows — where AI systems take autonomous actions rather than simply responding to prompts — create particularly complex infrastructure challenges, requiring tight coordination between specialized AI accelerators and general-purpose computing.

Ironwood is more than incremental improvement over Google's sixth-generation TPUs. According to technical specifications shared by the company, it delivers more than four times better performance for both training and inference workloads compared to its predecessor — gains that Google attributes to a system-level co-design approach rather than simply increasing transistor counts.

The architecture's most striking feature is its scale. A single Ironwood "pod" — a tightly integrated unit of TPU chips functioning as one supercomputer — can connect up to 9,216 individual chips through Google's proprietary Inter-Chip Interconnect network operating at 9.6 terabits per second. To put that bandwidth in perspective, it's roughly equivalent to downloading the entire Library of Congress in under two seconds.

This massive interconnect fabric allows the 9,216 chips to share access to 1.77 petabytes of High Bandwidth Memory — memory fast enough to keep pace with the chips' processing speeds. That's approximately 40,000 high-definition Blu-ray movies' worth of working memory, instantly accessible by thousands of processors simultaneously. "For context, that means Ironwood Pods can deliver 118x more FP8 ExaFLOPS versus the next closest competitor," Google stated in technical documentation.

The system employs Optical Circuit Switching technology that acts as a "dynamic, reconfigurable fabric." When individual components fail or require maintenance — inevitable at this scale — the OCS technology automatically reroutes data traffic around the interruption within milliseconds, allowing workloads to continue running without user-visible disruption.

This reliability focus reflects lessons learned from deploying five previous TPU generations. Google reported that its fleet-wide uptime for liquid-cooled systems has maintained approximately 99.999% availability since 2020 — equivalent to less than six minutes of downtime per year.

Perhaps the most significant external validation of Ironwood's capabilities comes from Anthropic's commitment to access up to one million TPU chips — a staggering figure in an industry where even clusters of 10,000 to 50,000 accelerators are considered massive.

"Anthropic and Google have a longstanding partnership and this latest expansion will help us continue to grow the compute we need to define the frontier of AI," said Krishna Rao, Anthropic's chief financial officer, in the official partnership agreement. "Our customers — from Fortune 500 companies to AI-native startups — depend on Claude for their most important work, and this expanded capacity ensures we can meet our exponentially growing demand."

According to a separate statement, Anthropic will have access to "well over a gigawatt of capacity coming online in 2026" — enough electricity to power a small city. The company specifically cited TPUs' "price-performance and efficiency" as key factors in the decision, along with "existing experience in training and serving its models with TPUs."

Industry analysts estimate that a commitment to access one million TPU chips, with associated infrastructure, networking, power, and cooling, likely represents a multi-year contract worth tens of billions of dollars — among the largest known cloud infrastructure commitments in history.

James Bradbury, Anthropic's head of compute, elaborated on the inference focus: "Ironwood's improvements in both inference performance and training scalability will help us scale efficiently while maintaining the speed and reliability our customers expect."

Alongside Ironwood, Google introduced expanded options for its Axion processor family — custom Arm-based CPUs designed for general-purpose workloads that support AI applications but don't require specialized accelerators.

The N4A instance type, now entering preview, targets what Google describes as "microservices, containerized applications, open-source databases, batch, data analytics, development environments, experimentation, data preparation and web serving jobs that make AI applications possible." The company claims N4A delivers up to 2X better price-performance than comparable current-generation x86-based virtual machines.

Google is also previewing C4A metal, its first bare-metal Arm instance, which provides dedicated physical servers for specialized workloads such as Android development, automotive systems, and software with strict licensing requirements.

The Axion strategy reflects a growing conviction that the future of computing infrastructure requires both specialized AI accelerators and highly efficient general-purpose processors. While a TPU handles the computationally intensive task of running an AI model, Axion-class processors manage data ingestion, preprocessing, application logic, API serving, and countless other tasks in a modern AI application stack.

Early customer results suggest the approach delivers measurable economic benefits. Vimeo reported observing "a 30% improvement in performance for our core transcoding workload compared to comparable x86 VMs" in initial N4A tests. ZoomInfo measured "a 60% improvement in price-performance" for data processing pipelines running on Java services, according to Sergei Koren, the company's chief infrastructure architect.

Hardware performance means little if developers cannot easily harness it. Google emphasized that Ironwood and Axion are integrated into what it calls AI Hypercomputer — "an integrated supercomputing system that brings together compute, networking, storage, and software to improve system-level performance and efficiency."

According to an October 2025 IDC Business Value Snapshot study, AI Hypercomputer customers achieved on average 353% three-year return on investment, 28% lower IT costs, and 55% more efficient IT teams.

Google disclosed several software enhancements designed to maximize Ironwood utilization. Google Kubernetes Engine now offers advanced maintenance and topology awareness for TPU clusters, enabling intelligent scheduling and highly resilient deployments. The company's open-source MaxText framework now supports advanced training techniques including Supervised Fine-Tuning and Generative Reinforcement Policy Optimization.

Perhaps most significant for production deployments, Google's Inference Gateway intelligently load-balances requests across model servers to optimize critical metrics. According to Google, it can reduce time-to-first-token latency by 96% and serving costs by up to 30% through techniques like prefix-cache-aware routing.

The Inference Gateway monitors key metrics including KV cache hits, GPU or TPU utilization, and request queue length, then routes incoming requests to the optimal replica. For conversational AI applications where multiple requests might share context, routing requests with shared prefixes to the same server instance can dramatically reduce redundant computation.

Behind these announcements lies a massive physical infrastructure challenge that Google addressed at the recent Open Compute Project EMEA Summit. The company disclosed that it's implementing +/-400 volt direct current power delivery capable of supporting up to one megawatt per rack — a tenfold increase from typical deployments.

"The AI era requires even greater power delivery capabilities," explained Madhusudan Iyengar and Amber Huffman, Google principal engineers, in an April 2025 blog post. "ML will require more than 500 kW per IT rack before 2030."

Google is collaborating with Meta and Microsoft to standardize electrical and mechanical interfaces for high-voltage DC distribution. The company selected 400 VDC specifically to leverage the supply chain established by electric vehicles, "for greater economies of scale, more efficient manufacturing, and improved quality and scale."

On cooling, Google revealed it will contribute its fifth-generation cooling distribution unit design to the Open Compute Project. The company has deployed liquid cooling "at GigaWatt scale across more than 2,000 TPU Pods in the past seven years" with fleet-wide availability of approximately 99.999%.

Water can transport approximately 4,000 times more heat per unit volume than air for a given temperature change — critical as individual AI accelerator chips increasingly dissipate 1,000 watts or more.

Google's announcements come as the AI infrastructure market reaches an inflection point. While Nvidia maintains overwhelming dominance in AI accelerators — holding an estimated 80-95% market share — cloud providers are increasingly investing in custom silicon to differentiate their offerings and improve unit economics.

Amazon Web Services pioneered this approach with Graviton Arm-based CPUs and Inferentia / Trainium AI chips. Microsoft has developed Cobalt processors and is reportedly working on AI accelerators. Google now offers the most comprehensive custom silicon portfolio among major cloud providers.

The strategy faces inherent challenges. Custom chip development requires enormous upfront investment — often billions of dollars. The software ecosystem for specialized accelerators lags behind Nvidia's CUDA platform, which benefits from 15+ years of developer tools. And rapid AI model architecture evolution creates risk that custom silicon optimized for today's models becomes less relevant as new techniques emerge.

Yet Google argues its approach delivers unique advantages. "This is how we built the first TPU ten years ago, which in turn unlocked the invention of the Transformer eight years ago — the very architecture that powers most of modern AI," the company noted, referring to the seminal "Attention Is All You Need" paper from Google researchers in 2017.

The argument is that tight integration — "model research, software, and hardware development under one roof" — enables optimizations impossible with off-the-shelf components.

Beyond Anthropic, several other customers provided early feedback. Lightricks, which develops creative AI tools, reported that early Ironwood testing "makes us highly enthusiastic" about creating "more nuanced, precise, and higher-fidelity image and video generation for our millions of global customers," said Yoav HaCohen, the company's research director.

Google's announcements raise questions that will play out over coming quarters. Can the industry sustain current infrastructure spending, with major AI companies collectively committing hundreds of billions of dollars? Will custom silicon prove economically superior to Nvidia GPUs? How will model architectures evolve?

For now, Google appears committed to a strategy that has defined the company for decades: building custom infrastructure to enable applications impossible on commodity hardware, then making that infrastructure available to customers who want similar capabilities without the capital investment.

As the AI industry transitions from research labs to production deployments serving billions of users, that infrastructure layer — the silicon, software, networking, power, and cooling that make it all run — may prove as important as the models themselves.

And if Anthropic's willingness to commit to accessing up to one million chips is any indication, Google's bet on custom silicon designed specifically for the age of inference may be paying off just as demand reaches its inflection point.

Presented by Arm

AI is no longer confined to the cloud or data centers. Increasingly, it’s running directly where data is created — in devices, sensors, and networks at the edge. This shift toward on-device intelligence is being driven by latency, privacy, and cost concerns that companies are confronting as they continue their investments in AI.

For leadership teams, the opportunity is clear, says Chris Bergey, SVP and GM, of Arm’s Client Business: Invest in AI-first platforms that complement cloud usage, deliver real-time responsiveness, and protect sensitive data.

"With the explosion of connected devices and the rise of IoT, edge AI provides a significant opportunity for organizations to gain a competitive edge through faster, more efficient AI," Bergey explains. "Those who move first aren’t just improving efficiency, they’re redefining what customers expect. AI is becoming a differentiator in trust, responsiveness, and innovation. The sooner a business makes AI central to its workflows, the faster it compounds that advantage."

Enterprises are discovering that edge AI isn’t just a performance boost — it’s a new operational model. Processing locally means less dependency on the cloud and faster, safer decision-making in real time.

For instance, a factory floor can analyze equipment data instantly to prevent downtime, while a hospital can run diagnostic models securely on-site. Retailers are deploying in-store analytics using vision systems while logistic companies are using on-device AI to optimize fleet operations.

Instead of sending vast data volumes to the cloud, organizations can analyze and act on insights where they emerge. The result is a more responsive, privacy-preserving, and cost-effective AI architecture.

Working with Alibaba’s Taobao team, the largest Chinese ecommerce platform, Arm (Nasdaq:Arm) enabled on-device product recommendations that update instantly without depending on the cloud. This helped online shoppers find what they need faster while keeping browsing data private.

Another example comes from consumer tech: Meta’s Ray-Ban smart glasses, which blend cloud and on-device AI. The glasses handle quick commands locally for faster responses, while heavier tasks like translation and visual recognition are processed in the cloud.

"Every major technology shift has created new ways to engage and monetize," Bergey says. "As AI capabilities and user expectations grow, more intelligence will need to move closer to the edge to deliver this kind of immediacy and trust that people now expect."

This shift is also taking place with the tools people use every day. Assistants like Microsoft Copilot and Google Gemini are blending cloud and on-device intelligence to bring generative AI closer to the user, delivering faster, more secure, and more context-aware experiences. That same principle applies across industries: the more intelligence you move safely and efficiently to the edge, the more responsive, private, and valuable your operations become.

The explosion of AI at the edge demands not only smarter chips but smarter infrastructure. By aligning compute power with workload demands, enterprises can reduce energy consumption while maintaining high performance. This balance of sustainability and scale is fast becoming a competitive differentiator.

"Compute needs, whether in the cloud or on-premises, will continue to rise sharply. The question becomes, how do you maximize value from that compute?" he said. "You can only do this by investing in compute platforms and software that scale with your AI ambitions. The real measure of progress is enterprise value creation, not raw efficiency metrics."

The rapid evolution of AI models, especially those powering edge inferencing, multimodal applications, and low-latency responses, demands not just smarter algorithms, but a foundation of highly performant, energy-efficient hardware. As workloads grow more diverse and distributed, legacy architectures designed for traditional workloads are no longer adequate.

The role of CPUs is evolving, and they now sit at the center of increasingly heterogenous systems that deliver advanced on-device AI experiences. Thanks to their flexibility, efficiency, and mature software support, modern CPUs can run everything from classic machine learning to complex generative AI workloads. When paired with accelerators such as NPUs or GPUs, they intelligently coordinate compute across the system — ensuring the right workload runs on the right engine for maximum performance and efficiency. The CPU continues to be the foundation that enables scalable, efficient AI everywhere.

Technologies like Arm’s Scalable Matrix Extension 2 (SME2) bring advanced matrix acceleration to Armv9 CPUs. Meanwhile, Arm KleidiAI, its intelligent software layer, is extensively integrated across leading frameworks to automatically boost performance for a wide range of AI workloads, from language models to speech recognition to computer vision, running on Arm-based edge devices — without needing developers to rewrite their code.

"These technologies ensure that AI frameworks can tap into the full performance of Arm-based systems without extra developer effort," he says. "It’s how we make AI both scalable and sustainable: by embedding intelligence into the foundation of modern compute, so innovation happens at the speed of software, not hardware cycles."

That democratization of compute power is also what will facilitate the next wave of intelligent, real-time experiences across the enterprise, not just in flagship products, but across entire device portfolios.

As AI moves from isolated pilots to full-scale deployment, the enterprises that succeed will be those that connect intelligence across every layer of infrastructure. Agentic AI systems will depend on this seamless integration — enabling autonomous processes that can reason, coordinate, and deliver value instantly.

"The pattern is familiar as in every disruptive wave, incumbents that move slowly risk being overtaken by new entrants," he says. "The companies that thrive will be the ones that wake up every morning asking how to make their organization AI-first. As with the rise of the internet and cloud computing, those who lean in and truly become AI-enabled will shape the next decade."

Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. For more information, contact sales@venturebeat.com.

Presented by Salesforce

Vibe coding — the fast-growing trend of using generative AI to spin up code from plain-language prompts — is quick, creative, and great for instant prototypes. But many argue that it's not cut out for building production-ready business apps with the security, governance, and trusted infrastructure that enterprises require. In other words, a few saved hours in development can mean a future full of security vulnerabilities, endless maintenance, and scalability headaches, says Mohith Shrivastava, principal developer advocate at Salesforce.

"For rapid experimentation, building minimum viable products, and tackling creative challenges, vibe coding is a game-changer," Shrivastava says. "However, that same speed and improvisational nature are exactly what makes its application in a professional, enterprise setting a topic of intense debate. And the skepticism from the developer community is 100% justified."

The excitement is all about speed: going from a rough idea to a working prototype in hours, not weeks, is a massive advantage. But as Shrivastava shared, developers have been vocal about the potential downsides.

"When you apply vibe coding indiscriminately to an entire application stack, you’re not just moving fast; you’re accumulating risk at an unprecedented rate," Shrivastava explains. "The cons are significant."

That includes potential security nightmares, as AI models don't typically take into consideration the company's specific security policies. They can easily introduce vulnerabilities like hardcoded secrets or use insecure, hallucinated packages. Then there’s the issue of what Shrivastava calls "spaghetti code on steroids," or verbose code that lacks a coherent architectural pattern, creating a mountain of technical debt.

Equally concerning is the illusion of progress: vibe coding may complete 80% of a feature in record time, but the remaining 20% — the edge cases, performance tuning, and compliance work — becomes exponentially harder.

But does this mean vibe coding has no place in the enterprise?

"The idea that you can just vibe your way to a complex, secure, and maintainable enterprise application is a dangerous fantasy," Shrivastava says. "But — the pros are undeniable if it's used correctly. The key is not to avoid vibe coding, but to apply it intelligently in your enterprise."

You can't, and you absolutely should not, vibe code your entire enterprise stack with just any generic tool, Shrivastava warns. But when paired with no-, low-, or pro-code tools that are built for the enterprise, many of the gaps can be addressed. An enterprise-grade vibe coding solution, for example, can automatically scan for security issues, flag performance bottlenecks, and provide a safety net.

It’s also critical to understand which parts of an application suit this approach — and which demand a higher level of trust and control. Shrivastava divides the stack into red and green zones to illustrate.

The green zone is the presentation layer, or the UI and UX. It’s ideal for vibe coding, where developers can move fast and iterate quickly without much risk. In contrast is the red zone, which covers the foundational pillars of an application, including business logic and data layers.

Developer expertise remains the foundation for effective and safe vibe coding. But developers can be amplified by AI tools and emerging agents that are grounded in business context, connected to real applications, integrations, and data flows.

"A generic AI agent can't grasp your company's unique processes, but a context-aware tool can act as a powerful pair programmer, helping a developer draft complex logic or model data with greater speed and accuracy," Shrivastava says. "It’s about making the expert developer more efficient, not trying to do their job for them."

Some areas will always be high risk for ungoverned AI — especially infrastructure and security. Letting a generic AI agent configure firewalls or Identity and Access Management [IAM] policies without oversight, Shrivastava warns, is a recipe for disaster. The solution isn’t to avoid the red zone entirely, but to approach it with the right tools — ones that embed governance, security, and context from the ground up.

"The winning strategy is clear: Vibe code the green zone for agility, approach the red zone by augmenting your developers with powerful, context-aware tools, and never, ever DIY your core infrastructure with AI," he says.

To harness the power of enterprise vibe coding, Salesforce developed Agentforce Vibes. This new vibe coding offering for the enterprise includes Agentforce, an autonomous AI agent built to collaborate like a pair programmer on the Salesforce Platform. It’s designed precisely to provide developers with the right tools for the job, covering both the green and red zones. For the green zone, it offers the speed and agility to rapidly build UIs and prototypes. But its true power lies in how it augments developers in the red zone.

"Enterprise vibe coding like Agentforce lets organizations take AI-assisted development to the organizational level, accelerating coding, testing, and deployment, while ensuring consistency, security, and performance," says Dan Fernandez, VP of product, developer services at Salesforce. "It's not about throwing away governance for speed; it’s about integrating AI into every stage of the application lifecycle to work smarter."

Because Agentforce Vibes’ tooling is deeply integrated with your business context on the platform, it can safely assist with business logic and data modeling. Most importantly, it operates on a trusted platform. Instead of a DIY approach — jury-rigging a generic AI agent to handle your networking — developers build on a foundation that has security and governance built in, so they can innovate safely, knowing the most critical layers of the stack are secure and compliant.

Agentforce Vibes users are now tapping the tool to build around 20 to 25% of their new code base, according to Salesforce data, and users are accepting around 1.2 million lines of agentic code per month. That includes companies like Coinbase, CGI, Grupo Globo, and one of the top five banks in the U.S., which is using Agentforce Vibes capabilities to develop production-ready apps faster.

Agentforce Vibes is part of a suite of tools in Agentforce 360 that span from no-code and low-code to pro-code development. These tools are together helping customers develop and deploy at speeds previously unheard of.

With the low-code Agent Builder in Agentforce, the Secret Escapes team was able to build, test, and launch their agent to support customer service in just two weeks, compared to the six months it had previously taken the company to build and train a bot.

With Agentforce, 1-800Accountant autonomously resolved 70% of customer chat engagements during tax week in 2025, without writing a line of code, using Salesforce’s low-code tools and AI assistance. Meanwhile, media company Grupo Globo deployed agents to identify subscribers at risk of lapsing, offer personalized upgrades, cross-sell, and convert non-subscribers. As a result, Agentforce boosted Globo’s retention rates by 22% in less than three months.

Enterprise tools show that disciplined engineering and creative experimentation can coexist — and that balance, Shrivastava says, is the key to lasting innovation.

"Vibe coding is not a fad, but it's also not a silver bullet that will replace disciplined software engineering," Shrivastava says. "The smart path forward is a hybrid approach where human software skills are augmented with agentic intelligence. This balanced approach is how you get the best of both worlds: radical innovation at the edge and unwavering stability at the core."

Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. For more information, contact sales@venturebeat.com.

By now, enterprises understand that retrieval augmented generation (RAG) allows applications and agents to find the best, most grounded information for queries. However, typical RAG setups could be an engineering challenge and also exhibit undesirable traits. 

To help solve this, Google released the File Search Tool on the Gemini API, a fully managed RAG system “that abstracts away the retrieval pipeline.” File Search removes much of the tool and application-gathering involved in setting up RAG pipelines, so engineers don’t need to stitch together things like storage solutions and embedding creators.  

This tool competes directly with enterprise RAG products from OpenAI, AWS and Microsoft, which also aim to simplify RAG architecture. Google, though, claims its offering requires less orchestration and is more standalone. 

“File Search provides a simple, integrated and scalable way to ground Gemini with your data, delivering responses that are more accurate, relevant and verifiable,” Google said in a blog post. 

Enterprises can access some features of File Search, such as storage and embedding generation, for free at query time. Users will begin paying for embeddings when these files are indexed at a fixed rate of $0.15 per 1 million tokens. 

Google’s Gemini Embedding model, which eventually became the top embedding model on the Massive Text Embedding Benchmark, powers File Search. 

Google said File Search works “by handling the complexities of RAG for you.” 

File Search manages file storage, chunking strategies and embeddings. Developers can invoke File Search within the existing generateContent API, which Google said makes the tool easier to adopt. 

File Search uses vector search to “understand the meaning and context of a user’s query.” Ideally, it will find the relevant information to answer a query from documents, even if the prompt contains inexact words. 

The feature has built-in citations that point to the specific parts of a document it used to generate answers, and also supports a variety of file formats. These include PDF, Docx, txt, JSON and “many common programming language file types," Google says.

Enterprises may have already begun building out a RAG pipeline as they lay the groundwork for their AI agents to actually tap the correct data and make informed decisions. 

Because RAG represents a key part of how enterprises maintain accuracy and tap into insights about their business, organizations must quickly have visibility into this pipeline. RAG can be an engineering pain because orchestrating multiple tools together can become complicated. 

Building “traditional” RAG pipelines means organizations must assemble and fine-tune a file ingestion and parsing program, including chunking, embedding generation and updates. They must then contract a vector database like Pinecone, determine its retrieval logic, and fit it all within a model’s context window. Additionally, they can, if desired, add source citations. 

File Search aims to streamline all of that, although competitor platforms offer similar features. OpenAI’s Assistants API allows developers to utilize a file search feature, guiding an agent to relevant documents for responses. AWS’s Bedrock unveiled a data automation managed service in December. 

While File Search stands similarly to these other platforms, Google’s offering abstracts all, rather than just some, elements of the RAG pipeline creation. 

Phaser Studio, the creator of AI-driven game generation platform Beam, said in Google’s blog that it used File Search to sift through its library of 3,000 files.

“File Search allows us to instantly surface the right material, whether that’s a code snippet for bullet patterns, genre templates or architectural guidance from our Phaser ‘brain’ corpus,” said Phaser CTO Richard Davey. “The result is ideas that once took days to prototype now become playable in minutes.”

Since the announcement, several users expressed interest in using the feature.

Your complete GTM stack, powered by AI

Discussion | Link

Keyword‑driven AI articles auto‑posted to your blog

Discussion | Link

Framer for your actual codebase

Discussion | Link

Skip migration and launch MCP with built-in Auth

Discussion | Link

AI Clone that scales your expertise

Discussion | Link

We’re thrilled to announce that Ankit Gupta is joining YC as our newest General Partner.

Ankit has already worked with dozens of YC founders as a visiting partner in recent batches. He has a rare blend of deep machine learning expertise and firsthand startup experience that make him

After 12+ years, 25 batches, and working with more than a thousand startups, Dalton Caldwell is transitioning to Partner Emeritus at YC.

Dalton joined YC in 2011 and has contributed in a number of roles over the years, most recently as a Managing Partner. During his time here, he worked

The country is currently facing the longest government shutdown in American history. What will it take for Congress to end it once and for all?

The Tesla CEO is on course to become the world’s first trillionaire.

Tariffs are causing chaos for many industries, including construction. Here’s how some companies handle the chaos.

The new NFIB monthly survey finds a third of business owners were unable to fill jobs in October, and are cutting back their recruiting plans.

We keep blaming the wrong things for what ails American health care. This is the real problem—and how to fix it.

As artificial intelligence enters its dating era, it has taken on an increasing number of roles: cupid, wingman, even romantic interest. 

Where once people’s biggest concern was being unfortunately catfished by old photos and flattering filters, now if a person seems too good to be true, well, they might not even be human at all.  

Hily’s Dating App T.R.U.T.H. report surveyed 1,559 U.S. daters and found 82% of Gen Z and 87% of Millennials are already turning to AI in their dating lives. Up to 95% also plan to use it in the future. 

Just as in traditional dating, there are some double standards at play. For Gen Z, 62% say they’d be turned off if they discovered their match was using AI during the talking stage, despite being happy to use it themselves. This number increased to 70% for Millennials. 

As one Gen Z dater on Hily said: “I’d be less attracted. It’s lazy and unnecessary. Demonstrates a lack of creativity and care.” Another added: “It feels less authentic—and also a lot less fun. Isn’t dating supposed to be fun, not an imitation of the weird arms race we see in work?”

The ick is understandable. It brings up questions like, if someone needs the help of AI to carry a conversation, are they even that interested? If they aren’t willing to put in the work, why should you? And ultimately, if you hit it off, are you falling for the person or the algorithm? 

The most common uses of AI are to generate a dating app bio based on information provided and to offer suggestions for conversations. “If it generated their bio, that’s fine—if it’s accurate,” one Millennial dater on Hily said. “But not if it made things up or used misleading photos.”

Daters are burnt out from endless swiping, ghosting and a transactional dating culture. “Chatfishing” is one way to optimize and outsource some of the emotional labor of finding love (who said romance was dead?). After all, most have enlisted the help of the group chat in crafting a risky text or handed over full control to a friend to spice up a conversation. 

Some chatfishers take things a step further, having ChatGPT conduct entire conversations. A 2025 study from Norton found six in 10 people who use dating apps believe they’ve encountered at least one conversation written by AI. 

It’s also getting harder and harder to tell if the person you’re messaging is human or bot. A 2025 preprint paper showed that human judges who spoke with both OpenAI’s model GPT-4.5 and a human simultaneously, mistook the AI for a human 73% of the time.

While AI can be an effective wingman when it comes to clinching a first date, when it’s finally time to meet face-to-face, problems arise. Of the daters Hily surveyed, 53% of Gen Z and 66% of Millennials say they’d feel less confident on a real-life date after using AI to communicate with a match. 

“People tend to turn towards AI due to the fear that they’re not good enough and need to outsource to be better liked and well received,” Dr. Sabrina Romanoff, relationship expert at Hily Dating App and Harvard-trained clinical psychologist, told Fast Company. “This has a paradoxical effect as AI often white washes over the personality traits, uniqueness, and small touches that make them special.”

And if your AI persona is smarter, wittier, and funnier than you, you are only setting yourself up for failure. 

Hello again, and thank you for reading Fast Company’s Plugged In.

In 2013, David Min came to Disney CEO Bob Iger with a big idea. Min, a founding partner at Disney’s investment arm, Steamboat Ventures, was now head of innovation for the entire company. He had concluded that something fundamental needed to be done about Disney’s relationship with the tech industry.

“We—meaning The Walt Disney Company—didn’t really have a very good reputation at the time for working with startups,” he remembers. Tech accelerators such as Y Combinator, 500 Startups, and Techstars were changing how high-potential concepts got their shot at becoming thriving businesses. Min thought Disney might learn something by investing in such an accelerator.

Iger’s take: That idea wasn’t big enough. “His response to me was like, ‘Why would we do that?—we should just do it ourselves,’” remembers Min.

So Disney did. The entertainment and media behemoth launched its own accelerator, partnered with Techstars to get it rolling, and gave it the most logical possible name: Disney Accelerator. In 2014, it unveiled its first cohort of 11 startups.

Eleven years later, corporate accelerators within large companies are no longer such a daring notion. Actually, they’re quite common. But Disney’s take on the idea has had time to grow well beyond its origin as an exercise in reputational repair.

On Wednesday, Disney Accelerator held its 2025 demo day on the Disney studio lot. The event served to introduce this year’s cohort of four startups: animation studio Animaj, microdrama producer DramaBox, 3D printer Haddy, and 3D projection company Liminal Space.

The Disney employees who gathered at the studio’s Main Theater to watch a video presentation about this year’s cohort and then mingle with this year’s startups and alumni companies in person came from across the company’s myriad enterprises, including movies, broadcasting, theme parks, cruise ships, consumer products, and beyond. They represented a fraction of the almost 600 staffers who now engage with the accelerator program year-round.

“We had people all the way from facilities and maintenance to the chairman of that division coming in for this one particular company,” says Disney Accelerator general manager Bonnie Rosen, whose résumé includes time at Techstars as well as a startup that was part of Disney’s 2015 cohort. “Those types of vertical conversations happen within each division.”

More than any other long-lived Hollywood titan, Disney prides itself on being innovative to its core. It’s an understandable badge of honor given that Walt Disney himself embraced advances such as the talkies, Technicolor, and TV as they came along, making each fundamental to the way his namesake company entertained the world. Today, it’s usually no mystery why Disney was intrigued by any given company among the 60-plus that have been through its accelerator program.

At demo day, for example, Liminal Space showed off its technology in the studio’s Stage One building, where the original Mouseketeers filmed The Mickey Mouse Club in the 1950s. It projects particularly crisp, vivid 3D video that can be viewed using simple polarized glasses. It can also be interactive: One of the demos involved The Guardians of the Galaxy’s Rocket Raccoon bantering with attendees. Liminal’s system isn’t currently in use at Disney’s parks, but it seems like a natural.

As for Animaj, it’s what Disney itself was in the beginning: a small but ambitious animation studio. Like Toonstar, which I recently profiled, the Paris-based company has built its own software platform that uses AI to help creators figure out which stories will resonate with audiences and then expedite the process of turning them into animation. In this case, they’re stories for little kids.

“The vision that we have with all of our properties is to turn them into global franchises with all the different layers,” Antoine Lhermitte, the company’s CTO, told me. “So we start on YouTube. Then we create a premium production [version] to sell to linear platforms, digital platforms like Netflix, Amazon Prime, Disney+, etc. Then we layer in consumer products, and then, if the traction continues, the idea is to go to theaters.“ Lhermitte says he’s hopeful the accelerator might lead to Animaj and Disney creating content together; the startup doesn’t want to be a service provider that just licenses its software to other studios.

Then there’s Haddy. At first blush, it might seem a bit of an outlier in the Disney Accelerator portfolio. The Florida-based company counts military, maritime, and furniture among the verticals it’s pursuing for its 3D printing technology, which tariffs have made newly enticing as a way to bring manufacturing back to the U.S. But the same factory that can crank out a 3D printed boat can also produce a full-size, real-world replica of King Louie’s throne from The Jungle Book—and has, as an experiment for Disney’s Imagineering theme-park designers. (It took about 20 hours to print.)

Haddy was already working with Disney when it was invited to join this year’s accelerator program. The company has networked with around 200 Disney executives as a result of this association, and has found that the experience redounds to the benefit of its other businesses, and vice versa. “You’re always learning,” says head of sales Erin Smith. “A boat that we print for Brunswick boats, for example, makes us more experienced and smarter when we print a boat for the Disney Jungle Cruise.”

The fact that Haddy is well down the path of applying its technology to fields not at all tied to its Disney association reflects the accelerator’s investment strategy, which has evolved over time. At first, it focused on early-stage startups and offered each one a standard $120,000 investment (the same figure once offered by Y Combinator). Eventually, however, Disney concluded that it was better off striking bespoke deals with growth-stage startups—ones whose future wouldn’t be overly skewed by Disney’s stake and the potential to sign up the company as a customer.

These further-along businesses “aren’t reliant on Disney for the health of their business development pipelines,” says Min. “Disney is a pillar of what they’re trying to accomplish, but it’s one of many things, and we encourage that.”

Which is not to say that even startups that are already booming can’t benefit from being well-connected at Disney. ElevenLabs is best known for its ability to turn real people’s voices into uncannily accurate synthesized speech. When it joined the accelerator’s 2024 program, it had fewer than 100 employees but was already a unicorn. Now it’s at 350 people and is still hiring, and the contacts it’s made within Disney remain valuable. “Sports, film, TV—we’re talking to all of them, because each of those divisions could use our product in so many different ways,” says head of partnerships Dustin Blank. “The conversations are always super interesting.”

In one case, the accelerator welcomed a company was already a venerated institution, an unorthodox arrangement that seemed to have worked out well for all involved. When Epic Games—the creator of Unreal and the game development platform based on it—joined the 2017 Disney Accelerator program, it was more than 25 years old and on the cusp of releasing something called Fortnite. The massive multiplayer game went on to truly epic success. In 2024, the two companies announced a partnership involving Disney taking a $1.5 billion stake in Epic and collaborating with it on new games based on Disney franchises.

Like anyone investing in startups, Disney aims to see a financial return from its accelerator’s portfolio. It also clearly sees the potential to apply some of the technologies it learns about to keep its many businesses growing. (CFO Hugh Johnston spoke as part of the demo day’s video, during a presentation that name-checked the company’s cofounder and original bean counter, Roy O. Disney, almost as often as his younger brother.) Min’s original goal of bolstering the company’s perception among techies remains crucial as well.

Yet another goal is allowing Disney to help shape the future of technology. Consider robotics, a hot topic at the moment that Rosen mentions when I ask her about emerging technologies that Disney Accelerator cares about (besides AI, of course). She notes the challenges a free-range Disney bot faces, such as safely weaving its way around theme-park visitors and food carts. But she also says the company might make a contribution to figuring out how to make robots more personable.

“It’s that personality part where Disney creatives are uniquely positioned to [initiate] a real momentum shift in how robotics are thought about,” she says. “Those are areas that are very exciting, and we wouldn’t look at them in the same way that the broader market is.”

Given that the company happens to have more than 60 years of experience in getting humans to bond with robots—dating to when Disneyland got its Enchanted Tiki Room and Mr. Lincoln first read the Gettysburg Address at the 1964 New York World’s Fair—it’s not an idle claim. And it’s one no mere stripling of a startup can match.

You’ve been reading Plugged In, Fast Company’s weekly tech newsletter from me, global technology editor Harry McCracken. If a friend or colleague forwarded this edition to you—or if you’re reading it on FastCompany.com—you can check out previous issues and sign up to get it yourself every Friday morning. I love hearing from you: Ping me at hmccracken@fastcompany.com with your feedback and ideas for future newsletters. I’m also on Bluesky, Mastodon, and Threads, and you can follow Plugged In on Flipboard.

Forget AI companions. This $249 AI ring lets you talk to yourself
Stream believes the future of AI might live on your finger. Read More →

New court docs put Sam Altman’s honesty in spotlight again
A lot is riding on the AI industry’s ability to apply AI productively and safely in business and personal life in the next decade. Trust is a major factor. Read More →

Nintendo wins suit against streamer who flaunted pirated games
Crossing the gaming giant can be an expensive proposition. Read More → 

 Paul Allen’s AI nonprofit unveils satellite data platform
Ai2 wants OlmoEarth to help nonprofits and governments without the resources for state-of-the-art geospatial AI work.  Read More → 

 Oops, I got emotionally attached to this $429 AI pet
Casio’s Moflin is perhaps the first AI ‘companion’ to deliver on its promise. Read More → 

 Here are the best mobile AI apps
Prompt, research, design, prototype, and more—all from your phone. Read More →

When fewer people belong to unions and unions have less power, the impact goes beyond wages and job security. Those changes can hurt public health and make people more unhappy.

We’re economists who research labor and health issues. Those are two of the main findings of studies that we have conducted.

In the first study on this topic that we published in 2023, we found that increasing levels of union membership tends to make working-class people happier.

We zeroed in on a question in the General Social Survey, which the University of Chicago makes available. It asks respondents to choose whether they are “very happy,” “somewhat happy” or “not at all happy” with their life.

We found that, from 1993 to 2018, when the share of workers in counties along the borders of states with and without right-to-work laws who belong to unions rose by 1 percentage point, the average level of happiness for low-income residents moved 15% closer toward being “very happy”—a seemingly modest but noticeable change.

Right-to-work laws let workers skip paying union dues when they’re employed by a company that has negotiated a contract with a labor union. In states without right-to-work laws, those dues are mandatory. As a result, right-to-work laws weaken unions’ ability to negotiate better working conditions and reduce the share of workers who belong to unions.

But a higher rate of union membership didn’t significantly affect the happiness of higher-income people.

The first right-to-work laws were adopted by states in the 1940s. After a long lull, the pace picked up around 2000. These laws were in force in 26 states as of late 2025.

Four of those states made the switch between 2001 and 2015: Oklahoma in 2001, Indiana in 2012, Michigan in 2012 and Wisconsin in 2015. We used data collected in these four states to conduct what is known in economics as an “event study”—a research method that provides before-and-after pictures of a significant change that affects large numbers of people.

Michigan repealed its right-to-work law in 2024, but our data is from 2001-2015, and Michigan became a right-to-work state during that period and remained one for the rest of that time.

In a related working paper that we plan to publish in an upcoming edition of an academic journal, we looked into other effects of right-to-work laws. Specifically, we investigated whether, as more states adopted those laws, the gradual decline in union strength those statutes produce was contributing to an increase in opioid overdoses.

We used a research technique called the synthetic control method to assess whether declining union power has affected the number of opioid overdoses.

We drew our data from a variety of sources, including the Treatment Episode Data Set, the Centers for Disease Control and Prevention’s Multiple Cause of Death database, the Census Bureau’s Current Population Survey, the union membership and coverage database, and the Bureau of Labor Statistics’ Survey of Occupational Injuries and Illness and Census of Fatal Occupational Injuries.

We found that both fatal and nonfatal opioid overdoses increased within six years of the enactment of right-to-work laws in all four of the states we studied.

We primarily found a connection between opioid overdoses and right-to-work laws among men and male teens between ages 16 and 64—making them of working age—with dangerous jobs, such as roofing or freight moving, and little job security. They were people who tend to feel more job stress because they don’t have control over their work tasks and schedules.

We didn’t observe those same results for women or deaths from non-opioid drugs, such as cocaine.

Lower levels of unionization are linked to weaker job security and reduced workplace protections, previous research has shown. Our work suggests these factors may play a role in increasing demand for opioids.

The share of U.S. workers who belong to unions has fallen by half in the past four decades, declining from just over 20% in 1983 to a little under 10% in 2024.

Because unions advocate for better and safer working conditions, they can raise wages and living standards for their members. Interestingly, some of these benefits can also extend to people who don’t belong to unions.

An opioid use disorder crisis has devastated communities across the U.S. for more than 25 years. The death toll from drug overdoses soared from 17,500 in 2000 to 105,000 in 2023. The number of overdose deaths did fall in 2024, to about 81,000, but it remains historically high. Most fatal drug overdoses since the crisis began have been caused by opioids.

Throughout this crisis, government policies have focused largely on reducing the supply of prescription opioids, such as OxyContin, and illegal opioids, especially fentanyl, distributed outside the health care system.

Despite successful interventions to shut down pill mills—clinics that prescribe opioids without a valid medical reason—and expand access to prevention and treatment, drug overdoses remain a leading cause of death.

And we believe that our findings support results from earlier studies that determined despair is not just an emotional or biological reaction—it can also be a response to social and economic conditions.

We are continuing to research the connections between union membership and public health. The next question we are working on is whether a decline in union membership can have a multigenerational impact, going beyond the workers employed today and affecting the lives of their children and grandchildren.

Samia Islam is a professor of economics at Boise State University and Kelly Chen is an associate professor of economics at Boise State University.

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Quantum computing insiders, investors, and skeptics have been waiting on an announcement from the Defense Advanced Research Projects Agency (DARPA) that has enormous implications for the future of the industry: the list of companies that have survived Stage A of the agency’s Quantum Benchmarking Initiative (QBI) and are advancing to Stage B. 

The QBI was launched in July 2024 to “rigorously verify and validate whether any quantum computing approach can achieve utility-scale operation” by 2033, according to DARPA. In essence, the QBI seeks to determine if a quantum computer technology is worth pursuing—if its benefits will be greater than the effort and resources it takes to pursue them. 

For a technology that could produce world-changing feats but remains far from maturity—and into which billions of investment dollars have been flowing in recent months—the QBI validation is profound.

The QBI’s first judgments, announced yesterday, reconfigure the competitive landscape, bolstering some powerful incumbents and boosting lesser-known players and outlier approaches. They also delivered a formidable gut punch to a couple of industry pioneers.

“QBI is not a competition to narrow the field to a few ‘winners.’ Rather, the aim is to evaluate each company’s approach on its own merits” DARPA said in a press release. The agency makes clear that multiple participants could “demonstrate a path to an industrially useful quantum computer,” or perhaps none of them. 

During Stage A of QBI, companies had six months to provide detailed technical concepts that showed feasibility for creating utility-scale quantum systems. The initial stage included 18 companies—IBM, Quantinuum, Atom Computing, Alice & Bob, IonQ, Rigetti, and Xanadu, among them. 

These companies, which were eligible for up to $1 million in funding, represent a variety of approaches to building a quantum computer, down to their technology for constructing qubits, the fundamental building blocks of a quantum computation.

The 11 companies that were selected to move onto Stage B are pursuing different systems. Two companies, Atom Computing and QuEra Computing, use “neutral atom” qubits. IonQ and Quantinuum both use so-called trapped ions. Several companies use silicon spin qubits. IBM is pursuing superconducting qubits. And Xanadu uses a photonics-based qubit. 

Now, moving into Stage B, they will be asked for a comprehensive research and development plan capable of realizing their quantum computer, with an assessment of the risks associated with the plan, and their mitigation approaches. In Stage C, companies will work with the government to verify and validate that their utility-scale quantum computer concept can be constructed as designed and operated as intended. 

DARPA anticipates that additional teams will advance through stages A, B, and C. And companies have entered the evaluation process on varying timelines—Google Quantum AI, for example, joined late—resulting in staggered advancement across stages. 

Two companies, Microsoft and PsiQuantum, have already advanced to Stage C in a separate but related DARPA initiative called Underexplored Systems for Utility-Scale Quantum Computing (US2QC).

“IBM’s progression to Stage B of DARPA’s Quantum Benchmarking Initiative is a firm validation of IBM’s approach to delivering a large-scale, fault-tolerant quantum computer,” said IBM’s director of research Jay Gambetta in a company press release. “As the industry advances, we look forward to working with DARPA as they continue an unbiased review of potential viable strategies across the field.” 

“It’s not that we got in that matters as much as making sure we didn’t not get in,” says Christian Weedbrook, CEO and founder of Toronto-based Xanadu Quantum Technologies. “A lot of late nights the team was pushing themselves, and I said, we don’t want a black mark across us now for getting the stage B. One of the great things about [the process] is that it really forced us in different teams to work harder together.” 

Earlier this week, Xanadu announced plans to go public via SPAC, which Weedbrook expects to happen in the first or second quarter of 2026. “This would be the only way, currently at least, that investors can invest in a photonics-based approach,” he says.

“Even if you are able to do months of diligence, nothing really beats the team of scientists and researchers from across the [Department of Defense] labs, the national labs, and the intelligence and universities ecosystem that DARPA has assembled to do this diligence,” says Prineha Narang, a professor of physical sciences and electrical and computer engineering at UCLA and a partner at deep-tech venture firm DCVC. “This is far more than any VC or anyone on the private capital side could do.”

Narang says that companies that didn’t make it to Stage B should take it “very seriously.” 

“It doesn’t mean that they don’t have an approach that could eventually work, but just that today they couldn’t articulate and present [that plan] to DARPA,” she adds. 

The Stage A contenders that didn’t make the cut include Alice & Bob, which uses innovative “cat qubits,” Atlantic Quantum, Hewlett Packard Enterprise, Oxford Ionics, and Rigetti Computing, which was founded in 2013 and is the first full-stack, universal pure-play quantum computing company.

On the other hand, Narang notes that a few companies that made it to Stage B have term sheets that call out the QBI outcome as a condition for the next round of funding. “Several investments and deals are now in flight,” she says. “[They are] happening, and happening faster than the usual pace of these.”

President Donald Trump is adjusting his messaging strategy to win over voters who are worried about the cost of living with plans to emphasize new tax breaks and show progress on fighting inflation.

The messaging is centered around affordability, and the push comes after inflation emerged as a major vulnerability for Trump and Republicans in Tuesday’s elections, in which voters overwhelmingly said the economy was their biggest concern.

Democrats took advantage of concerns about affordability to run up huge margins in the New Jersey and Virginia governor races, flipping what had been a strength for Trump in the 2024 presidential election into a vulnerability going into next year’s midterm elections.

White House officials and others familiar with their thinking requested anonymity to speak for this article in order to not get ahead of the president’s actions. They stressed that affordability has always been a priority for Trump, but the president plans to talk about it more, as he did Thursday when he announced that Eli Lilly and Novo Nordisk would reduce the price of their anti-obesity drugs.

“We are the ones that have done a great job on affordability, not the Democrats,” Trump said at an event in the Oval Office to announce the deal. “We just lost an election, they said, based on affordability. It’s a con job by the Democrats.”

The White House is keeping up a steady drumbeat of posts on social media about prices and deals for Thanksgiving dinner staples at retailers such as Walmart, Lidl, Aldi and Target.

“I don’t want to hear about the affordability, because right now, we’re much less,” Trump told reporters Thursday, arguing that things are much better for Americans with his party in charge.

“The only problem is the Republicans don’t talk about it,” he said.

As of now, the inflation outlook has worsened under Trump. Consumer prices in September increased at an annual rate of 3%, up from 2.3% in April, when the president first began to roll out substantial tariff hikes that suddenly burdened the economy with uncertainty. The AP Voter Poll showed the economy was the leading issue in Tuesday’s elections in New Jersey, Virginia, New York City and California.

Grocery prices continue to climb, and recently, electricity bills have emerged as a new worry. At the same time, the pace of job gains has slowed, plunging 23% from the pace a year ago.

The White House maintains a list of talking points about the economy, noting that the stock market has hit record highs multiple times and that the president is attracting foreign investment. Trump has emphasized that gasoline prices are coming down, and maintained that gasoline is averaging $2 a gallon, but AAA reported Thursday that the national average was $3.08, about two cents lower than a year ago.

“Americans are paying less for essentials like gas and eggs, and today the Administration inked yet another drug pricing deal to deliver unprecedented health care savings for everyday Americans,” said White House spokesman Kush Desai.

Trump gets briefed about the economy by Treasury Secretary Scott Bessent and other officials at least once a week and there are often daily discussions on tariffs, a senior White House official said, noting Trump is expected to do more domestic travel next year to make his case that he’s fixing affordability.

But critics say it will be hard for Trump to turn around public perceptions on affordability.

“He’s in real trouble and I think it’s bigger than just cost of living,” said Lindsay Owens, executive director of Groundwork Collaborative, a liberal economic advocacy group.

Owens noted that Trump has “lost his strength” as voters are increasingly doubtful about Trump’s economic leadership compared to Democrats, adding that the president doesn’t have the time to turn around public perceptions of him as he continues to pursue broad tariffs.

There will be new policies rolled out on affordability, a person familiar with the White House thinking said, declining to comment on what those would be. Trump on Thursday indicated there will be more deals coming on drug prices. Two other White House officials said messaging would change — but not policy.

A big part of the administration’s response on affordability will be educating people ahead of tax season about the role of Trump’s income tax cuts in any refunds they receive in April, the person familiar with planning said. Those cuts were part of the sprawling bill Republicans muscled through Congress in July.

This individual stressed that the key challenge is bringing prices down while simultaneously having wages increase, so that people can feel and see any progress.

There’s also a bet that the economy will be in a healthier place in six months. With Federal Reserve Chair Jerome Powell’s term ending in May, the White House anticipates the start of consistent cuts to the Fed’s benchmark interest rate. They expect inflation rates to cool and declines in the federal budget deficit to boost sentiment in the financial markets.

But the U.S. economy seldom cooperates with a president’s intentions, a lesson learned most recently by Trump’s predecessor, Democrat Joe Biden, who saw his popularity slump after inflation spiked to a four-decade high in June 2022.

The Trump administration maintains it’s simply working through an inflation challenge inherited from Biden, but new economic research indicates Trump has created his own inflation challenge through tariffs.

Since April, Harvard University economist Alberto Cavallo and his colleagues, Northwestern University’s Paola Llama and Universidad de San Andres’ Franco Vazquez, have been tracking the impact of the import taxes on consumer prices.

In an October paper, the economists found that the inflation rate would have been drastically lower at 2.2%, had it not been for Trump’s tariffs.

The administration maintains that tariffs have not contributed to inflation. They plan to make the case that the import taxes are helping the economy and dismiss criticisms of the import taxes as contributing to inflation as Democratic talking points.

The fate of Trump’s country-by-country tariffs is currently being decided by the Supreme Court, where justices at a Wednesday hearing seemed dubious over the administration’s claims that tariffs were essentially regulations and could be levied by a president without congressional approval. Trump has maintained at times that foreign countries pay the tariffs and not U.S. citizens, a claim he backed away from slightly Thursday.

“They might be paying something,” he said. “But when you take the overall impact, the Americans are gaining tremendously.”

Associated Press writers Will Weissert and Michelle L. Price contributed to this report.

—Josh Boak, Associated Press

Fellow first-time honourees also include Hydron Energy, Picketa Systems, Vivid Machines.

The post Carbonova, SenseNet, Xatoms among cleantech startups named to latest Foresight 50 first appeared on BetaKit.

US military-backed initiative aims to build useful quantum computer by 2033.

The post Nord Quantique, Xanadu, Photonic advance to next phase of DARPA quantum race first appeared on BetaKit.

While industry feels heard, tech leaders want more details on AI and sovereignty plans.

The post Canadian tech sees #Budget2025 as “about-face” from last government’s stance on business first appeared on BetaKit.

Virtual care provider buys New Brunswick platform to improve ADHD offerings.

The post Maple expands virtual mental health care through Beyond ADHD acquisition first appeared on BetaKit.

After executive shakeup, 1Password preps for AI era and potential IPO.

The post “We could be public today”: 1Password crosses $400-million USD ARR milestone first appeared on BetaKit.

This article is visible for CLUB members only. If you are already a member but don’t see the content of this article, please login here. If you’re not a CLUB member yet, but you’d like to read members-only content like this one, have unrestricted access to the site and benefit from many additional perks, you […]

The post Weekly funding round-up! All of the European startup funding rounds we tracked this week (Nov. 03-07) appeared first on EU-Startups.

Eastleigh-based Aquark Technologies, innovators in miniaturising deployable sensing and timing technologies based on cold atoms, today announced it has been awarded a €1.6 million (£1.4 million) contract by Innovate UK. The contract will facilitate the future deployment of Aquark’s cold atom-based atomic clock, AQlock, at a major telecommunications site in the UK. NPL and Purbrook […]

The post Aquark Technologies secures €1.6 million Innovate UK contract to advance deployable quantum timing systems appeared first on EU-Startups.

Anthropic, the San Francisco-based AI company behind Claude, has announced the opening of new offices in Paris and Munich as part of its accelerated expansion across Europe. Following a year of rapid growth in the EMEA region – where the company has tripled its headcount – these new hubs are set to bolster Anthropic’s presence […]

The post Anthropic expands from US with Paris and Munich hubs amid booming European growth appeared first on EU-Startups.

From autonomous agriculture to encrypted finance, satellite-enabled infrastructure underpins 18% of the United Kingdom’s GDP – an estimated €515 million (£454 billion) of economic activity. Yet, as the global space economy booms and European neighbours ramp up capital deployment into sovereign launch systems and orbital manufacturing capabilities, the UK is at risk of drifting in […]

The post €515 billion at stake – Why the UK can’t afford to miss the SpaceTech boom appeared first on EU-Startups.

I’m an ex-GM at Uber and am now building a startup in the UK, and I have one very painful question: Why is hiring for a startup in the UK so damn difficult? Within our entrepreneurial network, we often see the repercussions of poor hiring. Dying from bad hires is faster and easier than from […]

The post The corporate delusion: How to sabotage your own hiring appeared first on EU-Startups.

The post Motley raises $1.5M to build the world’s first AI business reporting platform appeared first on Seedcamp.

The post From the Forces to Investing into Deeptech, Dual-Use, and Defence appeared first on Seedcamp.

The post Andel raises $4.5M to help US employees get affordable medication appeared first on Seedcamp.

The post Umony announces $15M Series A and powerhouse Advisory Board appeared first on Seedcamp.

The post DaltonTx raises £4 million seed round to build adaptive, AI-enabled drug discovery platforms appeared first on Seedcamp.

Sovereignty has mattered since the invention of the nation state—defined by borders, laws, and taxes that apply within and without. While many

The post Reclaiming Control: Digital Sovereignty in 2025 appeared first on Gigaom.

I just returned from SAP Sapphire 2025 in Orlando, and while SAP painted a compelling vision of an AI-powered future, I couldn’t

The post SAP Sapphire 2025 appeared first on Gigaom.

Executive Briefing What Happened: A stealthy, persistent backdoor was discovered in over 16,000 Fortinet firewalls. This wasn’t a new vulnerability – it

The post When Patching Isn’t Enough appeared first on Gigaom.

Have you ever been in a group project where one person decided to take a shortcut, and suddenly, everyone ended up under

The post The EU’s AI Act appeared first on Gigaom.

The term “data fabric” is used across the tech industry, yet its definition and implementation can vary. I have seen this across

The post Demystifying data fabrics – bridging the gap between data sources and workloads appeared first on Gigaom.

For many founders, success doesn’t just come from having the right idea — it comes from communicating it well. That’s the premise behind Prezent, the Palo Alto-based AI-powered communications startup that has once again secured a spot on Inc.’s Power Partner Awards 2025, honoring the most trusted B2B partners helping small businesses thrive. The Inc. […]

The post Prezent named among Inc.’s Power Partners 2025, empowering founders through AI-powered storytelling appeared first on StartUp Beat.

South by Southwest (SXSW) has unveiled the first wave of programming for its 40th edition, scheduled for March 12-18, 2026. The festival, a cornerstone for creative and tech pioneers, announced over 250 sessions shaped by its community-driven PanelPicker process. This year’s lineup spotlights the evolving relationship between humanity and technology. While AI remains a dominant […]

The post SXSW Pitch 2026 calls on the world’s most innovative startups to present their disruptive ideas appeared first on StartUp Beat.

As mobile gaming continues to grow, developers and publishers face a persistent challenge: how to reach iOS users in Europe without being slowed down by restrictive App Store processes. Onside.io, the leading Distribution Enabler for the iOS ecosystem, and Playgama, a global web game hub with over 300 million monthly players, are tackling this challenge […]

The post Onside.io and Playgama Partner to Bring 300 Million Web Gamers to iOS Through EU-Compliant Alternative App Store appeared first on StartUp Beat.

The tourism industry in 2025 is booming. According to the World Travel & Tourism Council, global travel spending is set to soar past $11 trillion by the end of this year. However, this phenomenal demand also means that the environmental toll associated with tourism is rising. The future of the industry hinges on addressing its […]

The post The Future of Sustainable Travel Hits the Spotlight in Dubai this October  appeared first on StartUp Beat.

As AI keeps transforming the way software is built, most engineering teams still face a pressing challenge: how to set up meaningful measurement processes that distinguish value from risk.  With 90% of the developers today already using AI, adoption alone is meaningless without governance and measurement. Most teams also lack appropriate monitoring tools, leading to […]

The post Software engineering intelligence pioneer Waydev launches platform to measure AI impact  appeared first on StartUp Beat.

It’s a wrap! The latest Culture, Religion & Technology (CRT 2025) event was held this past Wednesday, October Read More

The post Culture, Religion & Technology, take III (CRT 2025) at Mar-a-Lago – It’s a wrap! first appeared on VatorNews.

Advsr AI Spotlight Highlighting notable AI moves at the intersection of strategic operating companies and emerging startups. Big Read More

The post Top AI news of the week #24: Advsr AI Spotlight first appeared on VatorNews.

Healthtech and edtech are two of the fastest growing sectors, with the healthtech market size to reach $3.1 Read More

The post Chegg cuts 45% of its workforce, Medline files for an IPO, Global Healthcare Opportunities raises over €2.5B first appeared on VatorNews.

Advsr AI Spotlight Highlighting notable AI moves at the intersection of strategic operating companies and emerging startups. Big Read More

The post Top AI news of the week #23: Advsr AI Spotlight first appeared on VatorNews.

Healthtech and edtech are two of the fastest growing sectors, with the healthtech market size to reach $3.1 Read More

The post Hologic acquired for $18.3B, Electra Therapeutics raises $183M, and Calm/Storm Ventures launches a new €30M fund first appeared on VatorNews.

Choosing the right legal structure for your startup is a critical decision that can impact your business’s future…

The post 18 Important Considerations When Choosing a Legal Structure for Your Startup appeared first on StartupNation.

Though the holiday season is the traditional time for spending, a feeling of caution, even dread, has formed…

The post How Small Businesses Can Win Holiday Shoppers Despite Inflation and Tariffs appeared first on StartupNation.

Every business fills some type of need within an industry, but some redefine it entirely. That’s what my…

The post 4 Strategies for Becoming a Category Creator appeared first on StartupNation.

Building a startup is tough. You’re juggling funding, product development, hiring, and numerous other priorities. It’s no wonder…

The post Why Consistent Branding Is the Secret Weapon for Startup Growth appeared first on StartupNation.

The idea that a business lives or dies depending solely on the personal qualities of its founder—innovativeness and…

The post 30 Days That Changed My Business: The Net-30 Lesson appeared first on StartupNation.