Submissions from simonwillison.net

		Claude Sonnet 4.5 is probably the best coding model in world (at least for now) (simonwillison.net)
		4 points by vismit2000 7 hours ago \| past \| discuss
		Claude Sonnet 4.5 is probably the "best coding model in the world", at least now (simonwillison.net)
		9 points by coloneltcb 15 hours ago \| past \| discuss
		CompileBench: Can AI Compile 22-year-old Code? (simonwillison.net)
		2 points by speckx 5 days ago \| past \| discuss
		I think "agent" may have a widely enough agreed upon useful definition (simonwillison.net)
		5 points by simonw 11 days ago \| past \| discuss
		GPT‑5-Codex and upgrades to Codex (simonwillison.net)
		57 points by amrrs 14 days ago \| past \| 15 comments
		Models Can Prompt Now (simonwillison.net)
		6 points by KolenCh 15 days ago \| past \| 3 comments
		Recreating the Apollo AI adoption rate chart with GPT-5, Python and Pyodide (simonwillison.net)
		3 points by vismit2000 20 days ago \| past
		Claude's new Code Interpreter review (simonwillison.net)
		22 points by vismit2000 20 days ago \| past \| 7 comments
		Geoffrey Huntle Is Cursed: Making a GenZ slang programming language with Claude (simonwillison.net)
		4 points by kposehn 20 days ago \| past
		My review of Claude’s new Code Interpreter, released under a confusing name (simonwillison.net)
		9 points by _mu 20 days ago \| past
		Recreating the Apollo AI adoption rate chart with GPT-5, Python and Pyodide (simonwillison.net)
		2 points by simonw 21 days ago \| past
		Load Llama-3.2 WebGPU in the browser from a local folder (simonwillison.net)
		3 points by indigodaddy 21 days ago \| past
		Google's new AI mode is good, actually (simonwillison.net)
		129 points by xnx 22 days ago \| past \| 66 comments
		GPT-5 Thinking in ChatGPT (a.k.a. Research Goblin) is shockingly good at search (simonwillison.net)
		9 points by thunderbong 23 days ago \| past \| 2 comments
		GPT-5 Thinking in ChatGPT (a.k.a. Research Goblin) is good at search (simonwillison.net)
		361 points by simonw 23 days ago \| past \| 255 comments
		Beyond Vibe Coding (simonwillison.net)
		4 points by edward 24 days ago \| past \| 1 comment
		Fs_usage: See filesystem activity on Mac OS X (simonwillison.net)
		1 point by fanf2 25 days ago \| past
		[flagged] simonw has vibe-coded 124 useful tools (simonwillison.net)
		17 points by doppp 25 days ago \| past \| 14 comments
		Claude Opus 4.1 and Opus 4 degraded quality (simonwillison.net)
		4 points by tosh 27 days ago \| past
		An LLM is a lossy encyclopedia (simonwillison.net)
		512 points by tosh 32 days ago \| past \| 343 comments
		Symbex: Search Python code for functions and classes, then pipe them into a LLM (simonwillison.net)
		2 points by auraham 32 days ago \| past
		Piloting Claude for Chrome (simonwillison.net)
		2 points by Shank 34 days ago \| past
		The Summer of Johann: prompt injections as far as the eye can see (simonwillison.net)
		3 points by jmvldz 35 days ago \| past
		Agentic Browser Security: Indirect Prompt Injection in Perplexity Comet (simonwillison.net)
		4 points by donohoe 35 days ago \| past
		AI vuln-a-day for the month of August (simonwillison.net)
		4 points by username223 44 days ago \| past \| 1 comment
		Open weight LLMs exhibit inconsistent performance across providers (simonwillison.net)
		5 points by mdp2021 45 days ago \| past
		GPT-5 has a hidden system prompt (simonwillison.net)
		9 points by gronky_ 45 days ago \| past \| 2 comments
		Open weight LLMs exhibit inconsistent performance across providers (simonwillison.net)
		6 points by neehao 45 days ago \| past
		codespaces-llm: template for GitHub Codespaces (simonwillison.net)
		2 points by indigodaddy 46 days ago \| past
		Qwen3-4B-Thinking: "This is art–pelicans don't ride bikes " (simonwillison.net)
		4 points by nnx 47 days ago \| past
		More