You write GPL code so improvements flow back, then AWS wraps it in a managed service and the license never fires because nothing is distributed. Worked exactly as designed and still failed. Fontana's right that static permissions can't solve dynamic power problems. I'm less sure about his "capacities" framework though, it would need governance structures most projects can barely imagine.

What about AGPL?

Adoption is the problem. Most projects won't touch AGPL because it scares away corporate contributors, so you end up choosing between a license companies will use but can exploit, or one they'll just avoid. Closed the loophole only on paper imo.

So? If the goal was to attract corporations you'd release under a BSD license.

The "biggest model that fits" instinct is just wrong now. Compact models routinely beat massive predecessors from 12 months ago. Scaling laws only reliably predict pre-training loss anyway, not how the model actually performs on your task. Dug into the research behind this: https://philippdubach.com/posts/the-most-expensive-assumptio...

These stories always focus on a vulnerable person, but what is the chatbot optimizing for?? It wants you to keep talking. Longer sessions, more training data, better engagement numbers. A therapist has a professional reason to make you need them less. ChatGPT has the exact opposite incentive baked in

Even if theres no business purpose, tge function of these models is to produce tokens. That alone skews the intelligence capabilities.

Good taxonomy but in practice these compose more than they compete I think..

Production agent stacks already layer MCP for tool access on top of RAG for context retrieval, with RLM-style orchestration wrapping the whole thing. The question nobody answered cleanly to me yet is which layer owns state. I wrote about this decomposition (1) and RAG's specific failure mode with sequential data (2).

(1) https://philippdubach.com/posts/dont-go-monolithic-the-agent...

(2) https://philippdubach.com/posts/beyond-vector-search-why-llm...

N=6 Phase 1, so this is purely a safety readout. But does the stem cell patch actually prevents hydrocephalus or just delays it.

lets see 35 patients in Phase 1/2a

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 be independent of the thing being audited.

I wrote about the layer split happening in agent tooling: https://philippdubach.com/posts/dont-go-monolithic-the-agent...

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 criminally charged for monetizing classified intelligence on event contracts.

Textbook Akerlof: without enforcement the adverse selection spiral kills these markets before they get a chance to prove useful. I tracked all three cases here: https://philippdubach.com/posts/the-absolute-insider-mess-of...

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 from across the Atlantic. Bert Hubert calls it “BTW-as-a-service.”

Europe talks a lot about digital sovereignty while outsourcing the plumbing. I looked at the same dynamic in payments, where the EuroPA alliance is at least building European rails instead of just complaining about it: https://philippdubach.com/posts/europes-24-trillion-payment-...