{{first_name | Leader}}, welcome back.
AI companies are no longer just competing on models. They're competing on politics, perception, and who gets to define the next generation of AI. These are today's updates.
Sam Altman wants to give Washington a piece of OpenAI
Meta says its next model matched GPT-5.5. It hasn't launched yet.
Fable 5 is back. This time with tighter guardrails.
Tools, resources, and a prompt to use when decisions escalate and consume leadership time. ⬇️
. Top News.
The relationship between AI labs and Washington has been getting more complicated with every major model release, every IPO filing, and every capability jump that attracts congressional attention.
Sam Altman is reportedly trying to get ahead of that. The proposal, still early and still informal, would have the US government take a 5 percent stake in OpenAI, with the broader idea potentially extending to Google, Anthropic, xAI, and Meta. The framing is sovereign-wealth-style: share the gains from the AI boom with the public rather than letting them concentrate entirely in private hands.
At an internal town hall, Meta's superintelligence chief Alexandr Wang told employees that Watermelon, the company's next flagship model currently still in training, has matched OpenAI's GPT-5.5 on key benchmarks. The model uses roughly 10 times more compute than Meta's previous model, Avocado.
Wang also said an update to Meta's current model, Muse Spark, is coming soon with stronger coding and agentic capabilities, and suggested Meta's coding model could reach Claude Opus-level performance in the near future.
The benchmark claim is worth taking seriously and with caution at the same time. Internal benchmarks at a town hall are not independent evaluations, and the model hasn't shipped yet. What the claim does confirm is that Meta is spending aggressively and closing ground faster than most people expected. When Watermelon actually ships, that's when the comparison becomes meaningful.
Anthropic announced that Claude Fable 5 is being redeployed globally starting July 1, following conversations with the US government and the lifting of earlier restrictions.
The redeployment comes with changes. New classifiers are now built in to block a broader range of cybersecurity-related tasks. Some routine coding and debugging tasks may temporarily fall back to Opus 4.8 while the new classifiers settle in.
Anthropic is also working with Amazon, Microsoft, and Google on a shared framework for evaluating jailbreak severity and coordinating responses to misuse across platforms.
. Signals.
Tools
Links
. Poll.
What benefit does AI bring to supply chain management?
. Market.
Funding
Venice raised $65M (Series A) at a $1B valuation.
TwelveLabs raised $100M (Series B) to build video superintelligence.
Roles In AI
VP of Demand Generation at Deepgram
Strategic Enterprise Account Executive, CMEG at Databricks
Socials
. Prompt of the Day.
Decision Escalation Matrix
When to use this?
When too many decisions escalate upward, leadership time gets consumed.
I’ll describe a business function or initiative below.
Build a Decision Escalation Matrix that clarifies:
Decisions teams should make independently
Decisions that require manager approval
Decisions that must escalate to leadership
Clear criteria for escalation
Examples of each decision category
Keep it simple and operational.
Context: [describe team, function, or initiative]P.S. Get more such prompts in the Prompting Playbook (free for you)
Stay curious, {{first_name | leaders}}
PS. We did something different for our previous issue, you can find it here. Reply to this email and let us know if we should explore this format or something else you’d like to know!