GenAI Daily - February 16, 2026: India AI Impact Summit Launches, Snorkel's $3M Benchmarks Initiative, Goodfire's $150M Series B

Top Story

The India AI Impact Summit 2026 opens today at Bharat Mandapam in New Delhi, running through February 20 and representing a shift from safety-focused summits to discussions on tangible AI impact, implementation, and governance.

The five-day event draws delegations from over 100 countries, including 15-20 heads of government, more than 50 ministers, and dozens of global industry leaders, with more than 700 sessions spanning policy, technology, and societal impact.

The summit is structured around three guiding pillars called Sutras (People, Planet, and Progress) that expand into seven domain-focused working groups, including discussions on frontier technologies like safe deployment of generative AI, global regulatory coordination, AI infrastructure, and scalable AI solutions for developing economies.

The India AI Impact Expo hosts 300+ exhibitors from 30 countries across a 70,000-square-metre exhibition space, showcasing real-world AI applications in industry, governance, healthcare, education, and sustainability.

Why it matters: The timing and scale signal a critical pivot from AI safety rhetoric to practical deployment - exactly what enterprise buyers need as they evaluate real-world AI solutions beyond pilot programs.

Key Developments

Snorkel AI Announces $3M Open Benchmarks Grants Initiative

Snorkel AI announced a $3 million Open Benchmarks Grant initiative targeting the evaluation gap in agentic AI, funding academic and open-source teams building next-generation benchmarks that better reflect how agents actually operate in production, as the launch addresses how agentic systems struggle with long-horizon tasks, complex toolchains, and real-world decision-making while current benchmarks fail to capture these failure modes meaningfully.

Through the initiative and new work on long-horizon tasks, world-modeling, non-stationary environments, and multi-artifact outputs, Snorkel calls for benchmarks that better mirror real-world agent workloads with nuanced rubrics that assess correctness, robustness, and human uplift - not just single-step accuracy.

(Snorkel AI)

Impact: Enterprise teams deploying AI agents can finally get realistic performance metrics instead of misleading leaderboard scores that don't translate to production success.

Goodfire Raises $150M Series B for AI Interpretability Platform

Goodfire raised $150 million in Series B funding at a $1.25 billion valuation to continue its research into understanding AI models, with the company's funding announced February 5, as it continues efforts in the field of interpretability.

The round was led by B Capital, with participation from Juniper Ventures, DFJ Growth, Salesforce Ventures, Menlo Ventures, Lightspeed Venture Partners, South Park Commons, Wing Venture Capital, Eric Schmidt, and others.

One application of Goodfire's methods reduced hallucinations by half in a large language model, betting that this approach will underpin a paradigm shift in how AI is built.

The company recently identified a novel class of Alzheimer's biomarkers by applying interpretability techniques to an epigenetic model built by Prima Mente - the first major finding in the natural sciences obtained from reverse-engineering a foundation model.

(Goodfire)

Impact: Enterprise AI deployments can now debug model behavior and reduce hallucinations systematically rather than hoping for black-box reliability.

DigitalOcean Survey: 40% of AI Agent Work Still Requires Human Review

DigitalOcean released findings from a survey of developers showing that while agentic AI adoption grew from 35% to 52% of companies implementing AI as core business strategy, productivity gains are rising with 53% reporting saved time from agents, but only 10% report having fully autonomous agents and 40% say humans still review agentic work, with almost half of respondents saying most of their budget now goes toward AI inferencing over training.

Impact: The data confirms what enterprise teams suspected - AI agents still need significant human oversight, helping set realistic expectations for deployment timelines.

Product Launches

• Apple Siri Overhaul:

Apple officially announced a completely reimagined, AI-powered version of Siri set to debut in 2026, with fundamental transformation seeing Siri transition into a context-aware assistant capable of "on-screen awareness" and seamless cross-app integration, powered by partnership with Google to utilize its 1.2 trillion parameter Gemini AI model running on Apple's Private Cloud Compute to maintain strict privacy standards, targeting March 2026 release alongside iOS 26.4.

(Apple Intelligence)

• NVIDIA Earth-2 Weather Models:

NVIDIA introduces Earth-2, an open family of AI models and tools for global and local weather forecasting that reduces compute time and cost.

(NVIDIA)

• Microsoft Maia 200 Chip:

Microsoft introduced Maia 200, a 3nm inference accelerator with 216GB HBM3e and large on-chip SRAM, aimed at lowering AI serving costs and power use.

Funding & Deals

• Goodfire - $150M Series B:

AI interpretability research lab using techniques to understand, learn from, and design models, working to reduce hallucinations and extract insights from foundation models for scientific discovery.

• Led by B Capital (Goodfire)

• Ricursive Intelligence - $300M Series A:

In January 2026, raised $300 million (adding to prior $35M) in Series A round targeting AI-capable chip solutions amid global supply concerns.

• Led by Lightspeed Venture Partners

• Upwind - $250M Series B:

Builds AI-enabled cloud application security, secured Series B led by Bessemer Venture Partners with funding to help expand its runtime-first platform amid surging demand for secure AI operations.

• Led by Bessemer Venture Partners

• Skild AI - $1.4B Series C:

Raised at $14 billion valuation with backers including SoftBank, NVIDIA, and Jeff Bezos, developing general-purpose foundation models for robotics across physical environments.

• Led by SoftBank

Tomorrow's Watch List

• OpenAI Frontier Platform Rollout:

Currently available to limited number of users with plans to roll out more generally in coming months, with enterprise customers including HP, Oracle, State Farm, and Uber already testing the agent management platform.

• India AI Impact Summit Sessions: Continue through February 20 with enterprise AI case studies and governance frameworks
• Apple iOS 26.4 Developer Beta:

Expected around the week of February 23, with public release likely in March or early April.

Related reading: Check out this week's [Deep Insights analysis] for strategic context on these developments.