7777777phil

Observe-only at the OS level is the right design! You can't trust the agent to report what it actually did. This is part of why I think monolithic agent platforms won't last. Auditing has to…

Neuromorphic chips have been 5 years away for 15 years now.. Nevertheless the Schultz dopamine-TD error convergence is one of the coolest results in neuroscience

This is the third major case in about three months. Before the Iran bets there was a suspected Google insider who made $1.15M on Polymarket (22 of 23 correct bets in 24 hours) and Israeli soldiers cri…

Love that the Norwegian Consumer Council just hired a guy whose job title is “shittificator” for a straight-faced explainer video.

The Netherlands just handed its VAT system to an American company.. Not just the software. The servers, the maintenance, all run remotely from the US. €1.5B a week in tax revenue as a managed service …

I went through 12 bank AI research reports (1) last week and power was the one thesis that held up no matter where you stand on the productivity question. Barclays projects US data center power hittin…

Pretty cool, playing with the guidance scale slider here taught me more than re-reading the DDPM paper did. Thanks for sharing!!

API prices dropped 97% in two years so the model layer is already a commodity. The question is which context layer actually sticks. The OpenClaw example in the article (400K lines to 4K) is a nice pro…

Japan holds roughly $5 trillion in foreign assets and JGB yields are rising fast, collapsing the spread that made holding Treasuries worthwhile. When the world's largest creditor starts rebalanci…

while we are debating alignment and guardrails while hundreds of millions of people pour their private thoughts into chatbots with zero data sovereignty.

Always the same pattern.. you build an identity verification layer to "protect children," then that layer is there for everything else. The UK's Online Safety Act already showed how thi…

Only regulation moves the needle. switching costs probably compound faster than frustration. Every quarter you stay on a platform, your exit cost goes up a little. Not enough to leave today, but enoug…

This lines up with something I keep coming back to. Sara Hooker's research shows compact models now outperform massive predecessors on many tasks, and scaling laws only reliably predict pre-train…

Making it OpenAI-compatible and handling GDPR by default is a good wedge. The hard part will be keeping up with model releases fast enough that enterprises don't just go direct.

Pretty sure I didn’t want to post that here. But then I got rate limited and upon coming out of rate limit jail blindly pasted this comment where my page reloaded - my bad should have been here: http…

Managers now requesting longer documents specifically to summarize with AI is almost funny, they've turned a discipline mechanism into a reading-avoidance tool. Most enterprise AI adoption has th…

77 suspicious positions across 60 wallets, 13 brand-new accounts appearing 40 hours before the browser launch. First confirmed case of a major tech company firing over prediction market trades. I wrot…

One thing worth adding: the repo market underneath all of this is roughly $12.6 trillion in daily exposures, about $700B larger than previous estimates. Since this is one of my favourite rabbit holes:…

The auth logic was literally inverted. Blocking people it should allow, allowing people it should block. Probably any human reviewer would catch that in seconds, but AI code generation optimizes for c…

Doubling progression-free survival (17.6 vs 7.4 months) is a large effect size for recurrent prostate cancer.

Netflix has the distribution already, it doesn't need legacy IP. Paramount is betting that combining DC, Harry Potter, GoT, Top Gun, and Star Trek under one roof creates enough gravity to survive…

The sharp results all came from pairing domain expertise with detailed AGENTS.md files. The impressive Rust output happened because someone who knows Rust was steering it. Vague prompts got mediocre o…

$730B pre-money for a company where each model is roughly 2x profitable on its own, but each next model costs 10x the last. The whole thing only works if scaling keeps delivering. Research (Sara Hooke…

The main thing i see here is that prompts never touch a third-party server. If you're in a regulated industry or just don't want proprietary context hitting an API, running inference on your…