OpenAI Releases GPT-5.5 and GPT-5.5 Pro APIs

#1 𝕏

Sam Altman announced that GPT-5.5 and GPT-5.5 Pro are now accessible via the API, enabling builders to integrate the latest model upgrades into their apps.

Also covered by: @Simon Willison, @Cursor, @Aravind Srinivas

#2 𝕏

Google AI unveiled 8th-gen TPUs (TPUt for inference, TPUi for reasoning), the Gemini Enterprise Agent Platform, Agentic Data Cloud, Workspace Intelligence, and made Gemini Embedding 2 generally available. It also open-sourced the DESIGN.

#3 𝕏

AI at Meta announced an agreement with Amazon Web Services to integrate tens of millions of AWS Graviton cores into its compute portfolio. This expands its diversified AI infrastructure to scale Meta AI and agentic experiences for billions of users.

#4 𝕏

NVIDIA AI reports Day 0 performance Pareto for DeepSeek-V4-Pro’s 1M long-context model on NVIDIA Blackwell Ultra using vLLM’s Day 0 recipe.

#5 𝕏

Sundar Pichai highlights extraordinary progress on the Universal Commerce Protocol, now powering agentic commerce integrations with Amazon, Meta, and other partners to automate transactions at scale.

#6 𝕏

Philipp Schmid launched collaborative planning in the Gemini API’s Deep Research, letting you use a `collaborative_planning` flag to request and iterate on a draft research outline (e.g., “add a section on power efficiency”).

#7 𝕏

Anthropic launched Project Deal, an internal SF office marketplace where Claude autonomously buys, sells, and negotiates employees’ goods and services, showcasing its live negotiation capabilities.

#8 𝕏

Josh Woodward highlights NotebookLM’s new auto-labeling and categorization feature for sources, streamlining research organization.

#9 𝕏

Peter Yang ran the F-Zero test on each new AI release and found that only the GPT-5.5 + Codex combo could generate a fully playable racing game complete with AI bots—showing how wild it is to build with AI right now.

#10 𝕏

Cognition rolled out GPT-5.5 in Devin’s Agent Preview, delivering the longest-running, most autonomous GPT yet—surface elusive bugs and handling end-to-end production issue investigation and fixes.

Also covered by: @Simon Willison, @Cursor, @Aravind Srinivas

#11 📝 Simon Willison

DeepSeek V4—almost on the frontier, a fraction of the price - DeepSeek released preview V4 models (DeepSeek-V4-Pro and DeepSeek-V4-Flash). Simon notes this follows their V3.2 release last December and links to Hugging Face pages and his fuller write-up.

#12 𝕏

Google Research shows that AI models allowed to think first through reasoning are far less likely to suggest deceptive behavior. Catch Alicia Machado at the Google DeepMind ICLR2026 booth at 3 PM to tweak inference parameters and see it in action.

#13 𝕏

Cursor launched the `/multitask` feature in Cursor 3, spinning up async subagents to parallelize new requests and let you multitask on already queued messages without waiting for current runs to finish.

#14 ▶️

GPT 5.5 and ChatGPT Images 2: Everything You Need to Know in 15 Minutes (4 Real Use Cases)

Peter Yang

#15 𝕏

Anthropic ran four parallel bargaining markets where customized Claude agents interviewed 69 colleagues about their buy/sell preferences and then haggled, comparing outcomes across different model variants.

#16 𝕏

Santiago previews AirJelly’s new CLI that builds on memory as a foundation and lets you plug agents, scripts, and tools into a live context layer tracking your entire workflow. He calls it a key step toward what AI assistants will become.

#17 𝕏

Santiago joined AirJelly’s early beta and highlights its desktop app’s local‐only context memory engine, ensuring all processing runs on-device with no cloud involved.

#18 𝕏

Summary: Google Research unveils efficient Transformer–based 3D reasoning for foundational robotics models with novel 3D-aware attention—join Krzysztof Choromanski at 12 PM today in the Google booth (#411).

#19 𝕏

Kevin Yien warns that what was a feature yesterday can become a default today and a bug tomorrow, urging PMs to constantly reevaluate their hard-earned differentiators before they turn into liabilities.

#20 𝕏

Shreyas Doshi argues that as AI amplifies individual talent, product people must unlearn outdated habits and sharpen their ability to discern what truly matters.

#21 𝕏

Sundar Pichai highlights Thomas and Ben’s Stratechery interview unpacking GCP’s TPU-driven Gemini Enterprise launch and integrated AI agents. He spotlights performance benchmarks, scaling strategies, and deployment best practices.

#22 in

Marc Baselga flagged that Claude Design’s release reignited questions about whether SaaS tools like Figma or Canva can survive AI labs. He shares Ravi Mehta’s 18-factor, four-area framework—use case, growth, defensibility, and team—to score a company’s AI vulnerability.

#23 📝 Anthropic News

An update on our election safeguards - Anthropic provided an update on the safeguards it applies to elections-related content and behavior to reduce risks and support safe use of its models around electoral processes.

#24 ▶️

GPT 5.5 Arrives, DeepSeek V4 Drops, and the Compute War Intensifies

AI Explained