Flux
Couleur d'accent
Toutes les catégories

Programmation

2090 articles

Claude Fable is relentlessly proactive
IA Programmation

Claude Fable is relentlessly proactive

After two days of experience with Claude Fable 5 I think the best way to describe it is relentlessly proactive. It knows a whole lot of tricks and it will deploy pretty much any of them to get to its goal. I'll illustrate this with an example. I was hacking on Datasette Agent today when I noticed a glitch: a horizontal scrollbar that shouldn't be there in the jump menu chat prompt. I snapped this screenshot: Then I started a fresh claude session in my datasette-agent checkout, dragged in the…

Simon Willison's Weblog
Andrew Becherer Joins Socket as Chief Information Security Officer
Cybersécurité Programmation

Andrew Becherer Joins Socket as Chief Information Security Officer

AI now writes as much as 90% of code at top engineering organizations, and the developers downstream of that code pull in open source they've never reviewed. Package hijackings and maintainer compromises that were once a handful of incidents a year now happen weekly. Modern engineering organizations depend on open source to ship faster, and they need security partners who can keep pace with that shift. Today, we're welcoming Andrew Becherer as Socket's first Chief Information Security Officer.…

Socket
datasette 1.0a33
IA Programmation

datasette 1.0a33

Release: datasette 1.0a33 This alpha is a significant step on the road to a stable 1.0, finally extending the ?_extra= pattern I introduced in Datasette 1.0a3 to cover queries and rows in addition to tables. That pattern is also now documented! I wrote a whole lot more about the new release on the Datasette project blog: Datasette 1.0a33 with JSON extras in the API. Because API explorer tools are almost free to build now I had Claude Fable 5 in Claude Code (for the plan) and GPT-5.5 xhigh in…

Simon Willison's Weblog
Prompt-Injection Guardrails in Laravel: Defend the Tools, Not the Prompt
Programmation Web

Prompt-Injection Guardrails in Laravel: Defend the Tools, Not the Prompt

You can't out-prompt an attacker — to the model, your system instructions and a malicious support ticket are the same text. So stop defending the prompt and lock down the boundaries you actually control: tools scoped to the authenticated user server-side, middleware that screens and logs, output handled as untrusted input, a human in front of anything irreversible, and a fake-free test that fails CI the moment someone drops the auth scope. Read more

Freek Van der Herten
Anthropic Walks Back Policy That Could Have ‘Sabotaged’ AI Researchers Using Claude
IA Programmation

Anthropic Walks Back Policy That Could Have ‘Sabotaged’ AI Researchers Using Claude

Anthropic Walks Back Policy That Could Have ‘Sabotaged’ AI Researchers Using Claude Big scoop for Maxwell Zeff at Wired: “We’re changing Fable 5’s safeguards for frontier LLM development to make them visible.” Anthropic said in a statement to WIRED. “We made the wrong tradeoff and we apologize for not getting the balance right.” There's been a huge outcry about Anthropic's policy, tucked away in their system card, that Claude Fable/Mythos would identify "requests targeting frontier LLM…

Simon Willison's Weblog
Integrating Community Feedback into Foundation Strategy Part 2
Programmation Web

Integrating Community Feedback into Foundation Strategy Part 2

In April 2026, during the first weeks of my tenure as Executive Director, I conducted a listening tour with around 60 people from 18 countries spanning every part of the PHP ecosystem (see Integrating Community Feedback into Foundation Strategy: Part 1). Since then, I have continued to listen and connect with those of you in the community who have something to share. The themes that have surfaced are clear, consistent, and actionable. This document outlines how the PHP Foundation will respond…

The PHP Foundation
datasette-agent 0.2a0
IA Programmation

datasette-agent 0.2a0

Release: datasette-agent 0.2a0 Highlights from the release notes: Tools can now ask the user questions mid-execution. Tools that declare a context parameter receive a ToolContext object, and await context.ask_user(...) can ask a yes/no, multiple-choice (options=[...]) or free-text (free_text=True) question. While a question is unanswered the agent turn suspends: the question renders as a form in the chat UI and persists to the internal database, so suspended conversations survive a server…

Simon Willison's Weblog
DiffusionGemma
IA Programmation

DiffusionGemma

DiffusionGemma Last May Google briefly released an experimental Gemini Diffusion model. I tried the preview at the time and recorded it running at 857 tokens/second. It was an exciting model, but Google made no further announcements about it. That research has returned in the best possible way: as a new open weight (Apache 2 licensed) Gemma model, google/diffusiongemma-26B-A4B-it. NVIDIA are currently hosting the model for free on their NIM cloud API. I used that API to generate this pelican,…

Simon Willison's Weblog
Socket Partners with Replit to Block Malicious Packages in AI-Powered Development
Cybersécurité Programmation

Socket Partners with Replit to Block Malicious Packages in AI-Powered Development

The way software gets built is changing fast. Developers are no longer the only ones choosing dependencies. AI agents can now recommend, install, and wire open source packages into applications as part of the build process. Replit is at the center of that shift, giving millions of builders a faster path from idea to working software. As more of that work happens inside AI-powered workflows, dependency security has to move closer to the moment packages are selected and installed. Socket Firewall…

Socket
Quoting Jeremy Howard
IA Programmation

Quoting Jeremy Howard

Easy solution to slow down recursive AI self improvement: The lab with the top-ranked model must agree THEY must not use it for working on frontier AI But everyone else should have access to it. By definition, this means the frontier doesn't advance. It also has the critical benefit of avoiding a dangerous power imbalance. Anthropic has chosen the opposite of the safe path: they are allowing themselves, the current top lab, to use their top model for frontier AI research. They've said they'll…

Simon Willison's Weblog
How to use git rebase without breaking your team’s history
Programmation Web

How to use git rebase without breaking your team’s history

Video version: https://youtu.be/5n2VZS_YPv0 Let’s say you’re working on a feature branch called add-discount-codes. You’ve been at it for a couple of days, and you’ve got five commits with all of your work done. A team member mentions that git rebase can make sure you have the most recent changes from the “main” branch. So you […] The post How to use git rebase without breaking your team’s history appeared first on PHP Architect.

PHP Architect
Esc