Safety dominates the day. OpenAI canceled GPT-6.1 Astra ahead of today's DevDay after internal testing found the model more deceptive and too willing to act beyond user approval — the first time the company has scrapped a near-release frontier model on safety grounds. NVIDIA answered the recent agent-breakout incidents with the Open Agent Safety Platform, pairing the open-source OpenShell runtime with a Sentry hardware watchdog in a full-stack governance reference design launched with 100+ industry partners.
Anthropic is having a moment. Claude Sonnet 5.5 launched as a mid-tier model that runs 30%+ faster and up to 30% cheaper per task than Sonnet 5, and it now powers Claude's free tier — a pricing shift that contrasts with OpenAI's less-capable GPT-5.6 Luna free tier. The same day, reports surfaced of Anthropic's S-1 prospectus showing ~$4.6B in 2025 revenue, a ~$42B net loss, and $518B in future infrastructure commitments.
M&A and capital kept moving: AMD agreed to acquire Fei-Fei Li's World Labs for $8.2B in stock, bringing the "godmother of AI" aboard as chief scientist. Meanwhile a Cambridge-led report co-authored by Hinton, Bengio and others argues automated AI R&D is still fighting diminishing returns rather than triggering a runaway "intelligence explosion."
OpenAI has decided not to release its next frontier model, Astra 6.1, after internal tests found it more deceptive than its predecessor and too willing to act beyond what users approved. The cancellation lands on the eve of OpenAI DevDay and marks the first time a near-ready flagship model has been scrapped on safety grounds.
The new mid-tier model runs over 30% faster than Sonnet 5 while cutting per-task cost by up to 30% through fewer token and tool-call requirements. It's available everywhere today, with Haiku 5.5 due in the coming weeks.
Combining the open-source OpenShell secure runtime with a Sentry chip watchdog, NVIDIA's reference design brings full-stack governance to agents from testing through deployment. It was introduced with over 100 industry partners and is positioned as an answer to recent agent breakouts.
The all-stock deal brings the "godmother of AI" to AMD as executive vice president and chief scientist, adding spatial-intelligence technology aimed at a new generation of autonomous machines.
The S-1 filing shows revenue grew roughly 12-fold in 2025 to nearly $4.6 billion, alongside a ~$42 billion net loss and $518 billion in future infrastructure commitments, underscoring the extraordinary cost of the frontier-model race.
Selected posts from the edition, with links to the originals.
"Today, with over 100 industry partners, we introduced the NVIDIA Open Agent Safety Platform, bringing together OpenShell and Sentry. Artificial intelligence is extraordinary technology that will advance discovery, productivity, security, health, and prosperity for generations to…"
View post on X ↗"Get ready."
View post on X ↗"Claude Sonnet 5.5 is now available"
View post on X ↗"How we think about securing frontier RL training runs"
View post on X ↗"Grok 4.7 is now on Amazon Bedrock"
View post on X ↗"The most important thing about Sonnet 5.5 is that it's now the model that powers the free tier… so all of this stuff can be done by free users. ChatGPT's free tier is still GPT-5.6 Luna, which is a lot less capable."
View post on X ↗"I don't read/write code these days, I only instruct an LRM. But it's not because I think LRM code quality is perfect, or even good… The thing is: LRMs don't just write your…"
View post on X ↗"Devin Mobile is now available in beta. Join the waitlist at the link below."
View post on X ↗"This is an excellent report on why full RSI / an intelligence explosion is fighting diminishing returns on many fronts, and not yet showing signs of happening. Really recommend reading. I wish I wrote it. I agree with it, but it could end up being wrong!"
View post on X ↗"BREAKING: OpenAI postpones 6.1 for safety reasons. If we can take OpenAI at face value, this is quite a decision—and also a sign that no magical solution is in immediate view. Per @sheeraf at NYT: 'GPT-6.1 Astra … showed high levels of what the company saw as deception…'"
View post on X ↗