THE 60-SECOND VERSION
● Pentagon blacklist of Anthropic struck down as unlawful retaliation. All directives to be rescinded.
● NBER: 89 percent of executives report no productivity effect from three years of AI.
● NVIDIA posted a record quarter, then paused cloud-financing deals for smaller AI operators.
● 2 days to the 31 August double deadline.
Court strikes down the Pentagon blacklist
U.S. District Judge Rita Lin ruled Thursday that the Defense Department's designation of Anthropic as a national security supply-chain risk violated the First Amendment through unlawful retaliation and the Fifth Amendment through inadequate due process, and ordered the government to rescind all directives against the company.
The dispute began when contract talks collapsed over Defense Department language permitting use of Claude for any lawful purpose. Anthropic declined over domestic surveillance and autonomous weapons; the Pentagon's position is that private companies should not constrain military action.
Lin's 59-page order pointed to an inconsistency in the government's own conduct: Hegseth had earlier threatened to invoke the Defense Production Act, which treats a supplier as essential rather than as a threat. A second Anthropic case remains pending in Washington D.C., where an appeals court ruled the other way in April.
Full breakdown: what the ruling found, and why it is not settled →
Nine in ten executives report no productivity gain
The National Bureau of Economic Research surveyed executives on three years of AI adoption. More than 90 percent reported no effect on employment at their own firm, and 89 percent reported no effect on productivity. Job cuts across the sector have not slowed.
That matches Linear telemetry published this month by a completely different method — teams using coding agents went from 21 weekly pull requests to 65 while total development time rose, because review scales with volume and lands on humans.
Where AI does show measurable gains, and what separates those cases →
Also today
- NVIDIA posted another record quarter, then paused the cloud-financing arrangements that had been supporting smaller AI operators.
- Forcepoint X-Labs demonstrated prompt injection through invisible email text — HTML at zero font size, white on white. A visible 537-character message delivered 1,009 characters to the model, and all ten runs produced manipulated summaries, including moving an invoice deadline. Recommendation: summarisers should extract only user-visible content and treat email as untrusted input.
- Meta changed its smart glasses software after users found that covering the recording indicator mid-capture bypassed the privacy cutoff.
Two days to the deadlines
| Date |
What happens |
| Aug 31 |
Claude Sonnet 5 goes $2 to $3 per million input, $10 to $15 output, plus a tokenizer change adding 10 to 35 percent tokens on code |
| Aug 31 |
GPT-5.4 and GPT-5.4 mini leave Codex for ChatGPT sign-in users. Still available via API key |
| Aug 31 |
kimi-k2.5 and moonshot-v1 sunset, migrate to kimi-k3 |
| Early-mid Sept |
Grok 4.7 window per Musk. Still no model ID at docs.x.ai |
| Oct 1 |
OpenAI vs Apple hearing |
| Oct 24 |
deepseek-chat and deepseek-reasoner deprecated |
| ~Nov 21 |
GPT-5.6 Sol promotional pricing ends, $4/$20 back to $5/$30 |
Monday is a repricing and a migration at once for anyone running coding workloads across both Anthropic and OpenAI. Model your token volume this weekend rather than discovering it in September.
Still open
- GLM-5.3 weights. Listed for 28 August after a two-week safety hold, with the flagship licence still unstated. What to check before downloading →
- Claude chat and Cowork share one memory, cloud-only, with no way to separate them. The controls and defaults →
- OpenAI's public S-1. Filed confidentially 8 June, still not on EDGAR.
FAQ
What did the judge rule about Anthropic and the Pentagon?
That the supply-chain risk designation violated the First Amendment through unlawful retaliation and the Fifth Amendment through inadequate due process, and that the government must rescind all directives against the company.
Does the Pentagon have to use Claude now?
No. The judge stated the Defense Department remains free to choose its AI vendor. The ruling struck down the designation, not the procurement decision.
What did the NBER survey find?
More than 90 percent of executives reported no employment effect at their own firm from three years of AI, and 89 percent reported no productivity effect.
What changes on 31 August?
Claude Sonnet 5 rises to $3 in and $15 out per million with a tokenizer change alongside, and GPT-5.4 and GPT-5.4 mini leave Codex for ChatGPT sign-in users. Both remain available via API.