Simon Willison
Independent developer; creator of Datasette and LLM
Turns experiments with coding agents into explanations of what to build, how to inspect the result and how to keep learning from code a model wrote.
Why read Simon Willison?
Willison is a useful first stop when a generated application appears to work and you want to understand what comes next. His writing separates prompt-driven experiments from the responsibilities of maintaining software, with specific attention to reviewing code, testing behavior and explaining how a system works.
Follow the essays with his Agentic Engineering Patterns guide. The linear walkthrough example shows how to ask an agent for a tour of a codebase, using commands and captured output to support the explanation. It gives a practical next step for turning a quick prototype into something you can understand and improve.
For agent security, start with Willison’s lethal trifecta: private data, untrusted content and a route for sending information out. Map each capability to the tools in a proposed workflow. His later discussion of the Agents Rule of Two adds state-changing actions and clarifies why a lower-risk configuration is still different from a guarantee of safety.
Selected work
-
Not all AI-assisted programming is vibe coding (but vibe coding rocks)
When is a quick AI-built experiment enough? Willison explains the original meaning of vibe coding, why it is useful for exploration and what changes when other people depend on your software.
Cite this selection
XFollowList, “Not all AI-assisted programming is vibe coding (but vibe coding rocks)” reading notes in the Simon Willison profile. https://xfollowlist.com/people/simon-willison#work-7649974b16a0
The linked work is the original source; this permalink identifies XFollowList’s reason to read it. Original work published .
Profile-source checking date and references · Dated reading map
-
Vibe engineering
What skills matter when agents write the code? A concrete account of planning, documentation, tests, review and preview environments as part of an engineer's work with agents.
Cite this selection
XFollowList, “Vibe engineering” reading notes in the Simon Willison profile. https://xfollowlist.com/people/simon-willison#work-077ccf7d939a
The linked work is the original source; this permalink identifies XFollowList’s reason to read it. Original work published .
Profile-source checking date and references · Dated reading map
-
Linear walkthroughs
How do you learn what your generated app does? Willison walks through asking an agent to explain a SwiftUI project, with source excerpts and command output recorded using his Showboat tool.
Cite this selection
XFollowList, “Linear walkthroughs” reading notes in the Simon Willison profile. https://xfollowlist.com/people/simon-willison#work-9eec1068750c
The linked work is the original source; this permalink identifies XFollowList’s reason to read it. Original publication date is unknown.
Profile-source checking date and references · Dated reading map
-
Large Language Models can run tools in your terminal with LLM 0.26
How does a model’s request become an executed tool call? Willison demonstrates Python functions, tool plugins, a read-only SQL example and tool-call logs. The dated examples require a terminal, Python and a supported model.
Cite this selection
XFollowList, “Large Language Models can run tools in your terminal with LLM 0.26” reading notes in the Simon Willison profile. https://xfollowlist.com/people/simon-willison#work-977a831af8ee
The linked work is the original source; this permalink identifies XFollowList’s reason to read it. Original work published .
Profile-source checking date and references · Dated reading map
-
Structured data extraction from unstructured content using LLM schemas
How do you turn articles or images into queryable records? Follow reusable schemas, saved responses, newline-delimited JSON and SQLite. The author explains that output-format reliability varies by model.
Cite this selection
XFollowList, “Structured data extraction from unstructured content using LLM schemas” reading notes in the Simon Willison profile. https://xfollowlist.com/people/simon-willison#work-733b90efc510
The linked work is the original source; this permalink identifies XFollowList’s reason to read it. Original work published .
Profile-source checking date and references · Dated reading map
-
The lethal trifecta for AI agents: private data, untrusted content, and external communication
Which combination lets an attacker steal data through an agent? His explanation connects private inputs, attacker-controlled content and outbound communication, making the capabilities you need to separate visible.
Cite this selection
XFollowList, “The lethal trifecta for AI agents: private data, untrusted content, and external communication” reading notes in the Simon Willison profile. https://xfollowlist.com/people/simon-willison#work-c11eb7058f35
The linked work is the original source; this permalink identifies XFollowList’s reason to read it. Original work published .
Profile-source checking date and references · Dated reading map
-
New prompt injection papers: Agents Rule of Two and The Attacker Moves Second
His analysis of other researchers’ proposals extends the data-leak model to state-changing tools. Read the updates explaining the difference between lower risk and safety; the cited papers retain their own authorship.
Cite this selection
XFollowList, “New prompt injection papers: Agents Rule of Two and The Attacker Moves Second” reading notes in the Simon Willison profile. https://xfollowlist.com/people/simon-willison#work-08ba944ec088
The linked work is the original source; this permalink identifies XFollowList’s reason to read it. Original work published .
Profile-source checking date and references · Dated reading map
Projects & roles
Sources & further reading
Profile sources checked . This describes checking these references, separately from the publication dates of selected works.
- His own current biography and disclosures describe his independent development work and link @simonw.Link to this source note
- The LLM project's documentation links Willison's original announcements and explains what the software does.Link to this source note
- His maintained guide connects the essays to practical chapters on testing, agents and understanding code.Link to this source note
- His original framing of the data-exfiltration capability combination, linked to concrete incident reports.Link to this source note
- His own commentary, including updates to the Rule of Two discussion; he is not credited as author of the underlying papers.Link to this source note
Cite this profile’s source check
XFollowList, Simon Willison profile sources. Sources checked 2026-10-09. https://xfollowlist.com/people/simon-willison#sources-checked