S
Simon Willison
simonw.substack.com · Personal blog · on HN since 2023
Simon Willison writes about AI tools, LLM agents, Claude and ChatGPT, and prompt injection risks, mirrored from his main blog.
What HN saysCommenters call him a prolific, helpful writer on AI; HN also cites his 'lethal trifecta' framing for agent security risks.
27HN people shared it
9linked it in comments
22stories submitted
92story points
3 months agolast shared
2023people per year2026
Most-shared posts
- 1The lethal trifecta for AI agents5 people · 4 points · first shared 2025 · discussion on HN
- 2Homepage3 people · first shared 2023
- 3First impressions of Claude Cowork, Anthropic's general agent2 people · 3 points · first shared 2026 · discussion on HN
- 4Embracing the parallel coding agent lifestyle2 people · 3 points · first shared 2025 · discussion on HN
- 5A new way to extract detailed transcripts from Claude Code2 people · 1 points · first shared 2025 · discussion on HN
- 6Feed2 people · first shared 2024
- 7GPT-5 Thinking in ChatGPT (a.k.a. Research Goblin) is shockingly good at search1 people · 52 points · first shared 2025 · 13 comments on HN
- 8Claude Opus 4.5, and why evaluating new LLMs is increasingly difficult1 people · 5 points · first shared 2025 · 1 comments on HN
- 9Feed a video to a vision LLM as a sequence of JPEG frames on the CLI1 people · 4 points · first shared 2025 · discussion on HN
- 10Image segmentation using Gemini 2.51 people · 4 points · first shared 2025 · discussion on HN
- 11First impressions of the new Amazon Nova LLMs1 people · 2 points · first shared 2024 · discussion on HN
- 12Multi-modal prompt injection image attacks against GPT-4V1 people · 2 points · first shared 2023 · discussion on HN
- 13Building software on top of Large Language Models1 people · 2 points · first shared 2025 · discussion on HN
- 14Trying out the new Gemini 2.5 model family1 people · 2 points · first shared 2025 · discussion on HN
- 15Prompt injections as far as the eye can see1 people · 1 points · first shared 2025 · discussion on HN
What HN people wrote when they linked it
tough · 2025-07-21
Simon's a prolific writer on AI
http://simonw.substack.com https://simonwillison.net
simonw · 2023-11-17
My approach is to publish to my own blog at https://simonwillison.net and then copy and paste content from that into a Substack newsletter at https://simonw.substack.com a few times a month.
It's been working really well.
Substack don't have an API, but they do support copy and paste - so I built myself a tool that assembles my blog content into rich text I can copy and paste straight into the Substack editor.
I wrote about how that works here: https://simonwillison.net/2023/Apr/4/substack-observable/
AdrienBrault · 2023-04-07
Not "HN-like", but I have found Simon Willison's blog/newsletter very helpful:
- https://simonwillison.net
- https://simonw.substack.com
hahahaa · 2026-08-05
> The victim uploads a file to Rovo that contains a hidden prompt injection
Yeah this attack is possible on all modern agentic systems.
* Access to your private data
* Exposure to untrusted content
* The ability to externally communicate in a way that could be used to steal your data
(https://simonw.substack.com/p/the-lethal-trifecta-for-ai-agents)
And blocking it wholesale reduces usefulness of the agent so it is a tradeoff.
medi8r · 2026-02-28
Also agents cannot trust any data whatsoever they add to their context.
This puts reading email for example as a risk.
Probably not impossible to create a worm that convinces a claw to forward it to every email address in that inbox.
And then exfiltrate all the emails.
Then do a bunch of password resets.
Then get root access to your claw.
But not just email. Github issues, wikipedia, HN etc. may be poisoned.
See https://simonw.substack.com/p/the-lethal-trifecta-for-ai-agents but there may be more trifectas than that in a claw driven future.
stevenhuang · 2026-01-23
simow wrote a tool that does this for Claude code
https://simonw.substack.com/p/a-new-way-to-extract-detailed-transcript…
bs7280 · 2026-01-15
Simon Willison just posted about using claude in fly.io's dev sandboxes. I have not tried it yet but it looks promising.
https://simonw.substack.com/p/first-impressions-of-claude-cowork
mellosouls · 2025-09-26
Original @simonw article here:
https://simonw.substack.com/p/the-lethal-trifecta-for-ai-agents
https://simonwillison.net/2025/Aug/9/bay-area-ai/
Discussed:
https://news.ycombinator.com/item?id=44846922
CharlesW · 2024-12-20
One tip: Every Substack has a feed, which is a nice alternative to having everything dropped in your email in-box. You just append "/feed" to the end of a newsletter URL, e.g. https://simonw.substack.com/feed.