Google Research introduces Gemini-SQL2 text-to-SQL API

#1 𝕏

Google AI launched Gemini 3.5 Live Translate for speech-to-speech translation, upgraded NotebookLM with agentic chat, advanced reasoning, and new output formats, and introduced DiffusionGemma, an experimental text diffusion model.

#2 𝕏

Google Research introduced Gemini-SQL2, a text-to-SQL capability powered by Gemini 3.1 Pro that achieves state-of-the-art performance on the BIRD benchmark. It converts natural language into execution-ready SQL queries with breakthrough accuracy.

#3 𝕏

NVIDIA AI released MiniMax M3, a long-context multimodal model for text, image, and video reasoning, now accessible via a free GPU-accelerated endpoint on NVIDIA Build.

#4 📝 Anthropic News

Statement on the US government directive to suspend access to Fable 5 and Mythos 5 - Anthropic says the US government issued an export-control directive at 5:21pm (ET) ordering suspension of all access to Fable 5 and Mythos 5 by any foreign national (including foreign-national Anthropic employees), forcing Anthropic to abruptly disable those two models for all customers while other Anthropic models remain unaffected. The company says the government’s concern is a narrow, non‑universal jailbreak demonstrated to reveal minor, previously known vulnerabilities (which Anthropic says are also discoverable in other public models), defends its "defense in depth" safeguards and 30‑day retention policy, and calls the order disproportionate and lacking transparent technical justification.

#5 𝕏

OpenAI now lets Go, Plus, Pro, and Business users save and later use one free Codex rate limit reset.

#6 𝕏

Aravind Srinivas released Plan Mode to all computer users, and will now ensure full clarification of requirements before crafting the plan and pulling in the right connectors.

#7 𝕏

Google DeepMind launched a three-month Robotics Accelerator with 15 European startups, granting them access to its AI stack, Gemini Robotics models, and hands-on support from its teams.

#8 📝 Anthropic Engineering

How we contain Claude across products - A featured post describing Anthropic's approach to containing Claude across multiple products, focusing on reducing the potential blast radius as agents become more capable. It shares lessons learned while building containment for claude.ai, Claude Code, and Cowork.

#9 𝕏

Santiago built an image search demo using embeddings and Oracle’s AI Database vector store, showcasing its new all-in-one data storage capabilities.

#10 📝 Simon Willison

OpenAI WebRTC Audio Session, now with document context - Announces an updated OpenAI WebRTC Audio Session tool that supports the newer GPT‑Realtime‑2 model and allows pasting large document context so the realtime audio model can discuss it in the browser.

#11 𝕏

Santiago demonstrates how to equip Claude Code with Apify actors to parse any website worldwide and highlights the newly added MCP connector support that unlocks even more integration possibilities.

#12 𝕏

Garry Tan toggled forceBlockStreamingForReasoning = resolvedReasoningLevel==="on" in OpenClaw to surface reasoning traces in Claude Fable 5, giving a mind-blowing “tale of the tape” view of the AI’s thought process.

#13 ▶️

How to Get ChatGPT to Recommend Your Business (Full GEO Tutorial)

Helena Liu

#14 ▶️

Anthropic begged the world to stop AI… then shipped this

Fireship

#15 𝕏

Philipp Schmid unveiled Agents’ Last Exam (ALE), a benchmark testing agents on 1,000+ real-world professional tasks across 55 industries with deterministic grading.

#16 𝕏

Google Research published findings showing that dermatology AI tools help laypeople more accurately name skin conditions and get clearer guidance on what next steps to take.

#17 𝕏

Teresa Torres warns that human moderation harbors a hidden flaw—teams enforce policies inconsistently across moderators, and most companies don’t even realize it. She points to ML-based analysis (podcast linked) as a way to detect and correct these gaps.

#18 𝕏

Julien Chaumond, Hugging Face announced that oMLX by @jundotkim now supports the standard Hugging Face cache model directory, making it a powerful MLX server for local AI deployments.

#19 𝕏

Philipp Schmid launched Agents’ Last Exam, an open benchmark platform (https://agents-last-exam.org/) for evaluating autonomous LLM-powered agents on rigorous, multi-step exam problems. He also published a detailed arXiv paper (2606.

#20 𝕏

Garry Tan warns that AI coding tools often just accelerate the build-out of rules, approvals and layers—“the same cage, assembled faster”—and urges PMs to use that speed to create genuinely new experiences, not calcified processes.

#21 📝 Simon Willison

Statement on the US government directive to suspend access to Fable 5 and Mythos 5 - Simon reacts to Anthropic's announcement that the US government issued an export control directive to suspend access to Fable 5 and Mythos 5, calls it "nuts", includes Anthropic's statement excerpt, notes he still had access briefly via claude.ai, and shows script logs demonstrating when API access was cut off.

#22 📝 Anthropic News

Results from the first Anthropic Public Record - Anthropic’s November–December 2025 nationally representative online survey of 51,993 Americans found 48% ranked curing diseases among their top three hopes for AI, while 64% feared AI-induced job loss, 56% feared cognitive dependency, and 52% feared misinformation. Over 70% supported government regulation (top priorities: privacy 56%, child safety 52%, liability 49%), 47% favored holding AI companies legally liable, 44% prioritized safety over growth, and just 15% said they trust AI companies to decide how AI is developed and used.

#23 𝕏

Clem 🤗, Hugging Face argues that because benchmarks average across varied queries, Opus 4.8—despite a lower overall score—can outperform Fable 5 on specific subsets (e.g., GPQA Diamond, AA-Omniscience), so using Opus 4.

#24 𝕏

Shreyas Doshi warns B2B product teams often chase every customer pain, overlooking that buyers only care enough to solve a handful of high-impact issues and will resist switching for lesser ones.

#25 𝕏

Shreyas Doshi counsels PMs to ditch clever-sounding jargon and focus on plain, truth-based descriptions of product ideas to sharpen insights, boost team alignment, and speed up decision-making.