
Braid · November 5 · 23 min
OpenAI Pledges to Skip Data Center NDAs After the Midterm Rebuke
0:00 · OpenAI answers the data center vote-23:08
transcript
show notes
Two days after voters rejected data center construction at the ballot box, OpenAI published a community blueprint that promises to avoid NDAs, protect ratepayers, and cut peak-hour power use, while the Wall Street Journal reports it discussed resetting its 2027 capacity. Today's episode asks which of those promises bind anyone, and then covers a run of other news.
- OpenAI's blueprint introduces a National Community Compact and Compute Campus Policy Blueprint; CNBC and Bloomberg via Techmeme list the pledges, which matters because a proposal is not a contract.
- Axios has Chris Lehane's 1996 comparison, the community-payment race among Amazon, Microsoft and OpenAI, and the warning that new state leaders want binding agreements; the Wall Street Journal via Techmeme reports OpenAI discussed resetting its 2027 capacity plans.
- The Guardian and TechCrunch cover the political side as a Democratic Congress prepares to take committee control.
- Reuters, Bloomberg and Bloomberg on CoreWeave show hyperscalers up and developers and neocloud creditors down; the Financial Times and Reuters report Microsoft passing on a bigger Nscale option and Google backstopping TeraWulf.
- OpenAI's kernel post reports 948 Linux privilege-escalation bugs found and patched in one run; Smol AI's digest supplies the 1.2 million dollar cost, and SiliconANGLE shows the patch-cycle pressure on database vendors.
- Conductor's announcement, SiliconANGLE on Pierre, Business Insider on Cat Wu and the Wall Street Journal on Brockman show the coding-agent layer being absorbed by labs.
- CNBC reports Moonshot seeking Blackwell capacity from European neoclouds, and Geopolitechs reports DeepSeek's plan to pretrain V5 on Huawei Ascend.
- OpenAI's Astra report and Anthropic's provable inference post bracket the question of how far to trust a model's output.
- Ripple-65B and Clement Delangue's Wired interview cover a US-origin open-weight model and the argument that closed labs carry the risk.
- Wise's job agreement talk and Ramp's specialist-model talk set a bounded-authority design and a build-your-own-model result beside Blackstone's wait-for-the-next-model view.
- TechCrunch reports the backlash to Google's plan to cut off Search for schools that restrict AI.
links29





