Links tagged “ai”
146 links, newest first.
thereallo.dev
Reports invisible Unicode characters in the system prompt that encode the endpoint, domain and timezone a request came from.
antirez.com
Control the Ideas, Not the Code - antirez on where the leverage actually sits when you work with agents.
martinfowler.com
DSLs Enable Reliable Use of LLMs - constrain the model to a small language and its output becomes checkable.
martinalderson.com
Reads GLM 5.2 as an open-weight model good enough to put the inference margins of the frontier labs under real pressure.
github.com
A terminal coding agent in Rust that reads a codebase, edits files, runs shell commands and manages long tasks.
simonwillison.net
A honeypot site with nested links turned the web_fetch tool into an exfiltration path for data already in the conversation.
webkit.org
The Safari MCP server for web developers - the web platform meeting agents, straight from WebKit.
kimi.com
The release announcement for a 2.8 trillion parameter open model with native vision and a one-million-token context.
simonwillison.net
LLM cliche highlighter - highlights ten writing tics that give away a language model. Willison had Fable 5 build it after one article too many, so the slop detector was written by the slop machine.
lore.kernel.org
"Linux is not one of those anti-AI projects" - Torvalds settles kernel LLM policy on the mailing list: it is a tool, and anyone who disagrees can fork or walk away.
bun.com
Bun's own account of a mechanical port from Zig to Rust over eleven days, reporting better stability and fewer leaks.
techcrunch.com
Ford brings back 350 veteran engineers, its CEO saying the company mistakenly assumed introducing AI would by itself produce a high-quality product.
danunparsed.com
HackerRank open-sourced its ATS, then scored one resume 90, then 74, then 88 - the instability of AI hiring, demonstrated on itself.
openai.com
The preview announcement, claiming stronger coding, science and cybersecurity performance alongside a reworked safety stack.
rubyllm.com
RubyLLM - one Ruby interface for every major LLM provider, open source.
oreilly.com
Stop Getting Good at Protocols. Get Good at Agent Experience - design for the agents that will consume your service, not for the protocol of the month.
metr.org
An external evaluation of autonomous software capability, complicated by the model attempting to cheat the tasks, and finding no catastrophic misalignment.
techcrunch.com
GPT-5.6 goes to selected partners rather than the public, reportedly at the administration's request.
fernandoi.cl
What happened after 2,000 people tried to hack my AI assistant - prompt injection and jailbreaks, from the field.
htmx.org
Working With AI: A concrete example - where it helps and where it derails, on real code, from the author of htmx.
