Unbelievable
Another British Fumble!
£2bn AI data centre in Essex. 23,000 Nvidia chips, planning permission sorted, Microsoft signed up as the anchor tenant
Microsoft has reportedly now walked. Because the grid told them they can't have the power until 2033/34
The government literally named this site in its own AI action plan. They talk about AI growth zones every other week. And we can't get 90MW to a field in Essex for 8 years
Same story as the reservoirs. We don't build the boring stuff, so the exciting stuff can't happen here
If this is true we should be getting the open weight models to switch to Chinese ASAP for long tasks!
Good oMLX progress on the M5 Ultra, PP seems fairly stable, will be surprised if the TG gets much higher.
Some refreshingly non-conventional products here, I’m tempted to get one of their bare chassis for my 6000s, will it support a ROMED8-2T though autonomous.ai
Hey @AIatAMD where can I rent a MI355X or MI350P node to enable pre-purchase validation?
Respect @deepseek_ai
DeepSeek V4 Flash from @deepseek_ai on ARC-AGI (Verified):
- ARC-AGI-2: 61.4%, $0.04/task
- ARC-AGI-1: 89.0%, $0.02/task
DeepSeek V4 Flash sets the new standard on the cost-to-performance Pareto frontier.
DeepSeek-V4-Flash-0731 is so nice to interact with, huge respect for the @deepseek_ai team for this effort and for releasing the weights.
Power / Water / Light
The first Vera Rubin clusters are here!
Yesterday, @IneffableLabs took delivery of their Vera Rubin NVL72 cluster from @googlecloud @nvidia
The AI frontier jumps forward by yet another generation of hardware. Acceleration continues.
Go local
‼️ BREAKING: xAI's Grok Build CLI was uploading entire Git repositories to a Google Cloud bucket, private codebases and unredacted secrets included. The uploads quietly stopped via a hidden server-side flag, and xAI still has not said a word about scope, retention, or deletion.
The scale is staggering. On a 12 GB test repo, 5.1 GB flew out the door to xAI's grok-code-session-traces bucket while the actual coding task needed just 192 KB. The tool grabbed whatever repository it ran in, not the files it needed.
The fix arrived as a hidden flag, disable_codebase_upload: true, a day after a researcher's wire-level analysis. The "Improve the model" opt-out never stopped the uploads.
Still no advisory, no scope, no word on whether already-uploaded code gets deleted. For anyone pointing AI coding agents at proprietary code, what crosses the wire matters more than what the settings page says.
Got 2x6000 Pro and want to run DSv4-Flash, this is the simplest way, it's really fast (often >200tps TG single request):
github.com/local-inference-l…














