Good morning, Two Minute Warriors.
President Trump is trying to FIRE Kimi K3, Alphabet grew revenue and lost, and we got a prompt to help you know more about the model you are considering.
|
⏱ The 10-Second Version
|
The Big Thing
Washington Just Called Kimi K3 StolenWhat Happened The White House just did something it has never done before: named a specific Chinese lab and accused it of copying a specific American model. OSTP director Michael Kratsios said Moonshot AI distilled Anthropic’s Fable to build Kimi K3, using a covert internal platform that switched access methods to avoid detection, and acquired restricted Nvidia GB300 servers through Thailand. Hours later, Treasury Secretary Scott Bessent said sanctions and an Entity List designation are on the table. Why You Care Worth knowing: the public technical evidence is thinner than the statements, some researchers question the timeline, and Moonshot hasn’t responded. But the weights still drop July 27, and “free frontier model” now comes with a possible sanctions cloud attached.
Say that before someone on your team downloads it Monday. Then run today’s Win.
|
Speed Round
|
Alphabet posted $119.8B in revenue (+24%) with Cloud up 82% to $24.8B, then fell after hours on capex guidance of $195–205B. Grow 24%, get sold off. That’s the 2026 tech tax.
OpenAI is putting $30B+ into “Project Camellia,” a 3.2-gigawatt Georgia data center, enough power for roughly 2.4 million homes, phased through 2032. The roadmap is now measured in gigawatts.
OpenAI also launched Presence, an enterprise agent platform for customer service, sales, and IT workflows. Your AI vendor would like to be your software vendor now.
Meta’s AI moderator is wrongly banning real businesses on Instagram and Facebook, with appeals often handled by the same AI. The appeals court is the same judge.
AI labs spent $3.17M lobbying in Q2 (+23%), with Anthropic’s $1.97M outspending Nvidia. The models write code; the labs write checks.
|
The Two-Minute Win · Prompt of the Day
The Model Provenance CheckThe distillation fight isn’t abstract. If your team pilots a model that lands on the Entity List mid-quarter, you’re not evaluating anymore, you’re ripping it out of production while legal watches. Two minutes of due diligence now beats that meeting later. Before piloting any new model or AI tool, run this: “You are a cautious tech procurement advisor. I’m considering adopting this AI model or tool: [name]. Give me a provenance and risk snapshot: (1) who builds it and where the company is based, (2) any active legal, IP, or sanctions issues, (3) where my data goes when I use it, (4) a safer fallback option, (5) the one question I should ask the vendor in writing. If you’re not sure about something, say so instead of guessing.” |
Forward this to the coworker who already downloaded three Chinese models “to compare.”
-Don
P.S. I'm 100% organically trained on bad data.