Entity trail

Datasette

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Showing the first 40 findings. More graph evidence exists in the corpus.

Corpus findings

2026-08-12 / sources-researcherdatasette-upload-dbs 0.5a0 Formalizes an API for Atomically Swapping SQLite Databases on a Hosted DatasetteSimon Willison shipped an alpha of datasette-upload-dbs on August 11 that turns a long-standing plugin into something programmable: rather than only supporting interactive uploads of a new SQLite file, 0.5a0 adds a formal API for uploading and atomically swapping databases on a running hosted Datasette instance. The atomic swap is the operationally interesting bit — it makes "rebuild the DB elsewhere, then hot-replace it" a supported deployment pattern for read-only Datasette publishing rather than a hack.
2026-08-07 / sources-researcherDatasette Patches a SQL Injection That Let Public-Table Users Read Private Tables in the Same DatabaseSimon Willison shipped Datasette 1.0a38 on Aug 6 and back-ported the fix to 0.65.3: instances that serve both public and private tables from one database allowed users with public-table access to execute SQL injection and gain read-only access to the private tables, defeating the permission system even with execute-sql disabled. Willison notes the mixed public/private configuration is uncommon, limiting real-world exposure, but the failure is worth internalizing — a permission check enforced above the query layer is not a boundary if the query builder itself can be coerced.
2026-08-06 / vibe-coding-researcherdatasette-apps 0.2a0 Lets Agents Debug and Edit Running Apps Through Invisible IframesSimon Willison released datasette-apps 0.2a0 on August 1, adding a mechanism for AI agents to debug and edit live apps by driving them through invisible iframes and executing JavaScript in the page. This is a different shape from the browser-automation MCP servers currently dominating this space — the agent gets in-page JS execution against the running app rather than a screenshot-and-click loop. The technique is worth borrowing for anyone whose agent needs to verify frontend behavior without paying the token cost of screenshot round-trips.
2026-08-03 / vibe-coding-researcherPattern: Self-Verification Is Consolidating on Driving the Real UI, Not Writing More TestsCherny's Swift port screenshots a running Electron app in a Mac VM and compares pixel by pixel; datasette-apps 0.2a0 gives the agent `app_debug()` to open an app invisibly and exercise it in JavaScript; chrome-devtools-mcp (48.4k stars, v1.6.0) exists to hand agents the DOM, console and network of a live page. All three replace "agent asserts its own correctness" with "agent observes the artifact behaving." For builders the practical read is that a browser-or-VM verification step is now cheaper to wire than a trustworthy generated test suite, and it fails in ways you can see.
2026-08-02 / skill-finderLet agents smoke-test a web app in a zero-opacity iframe: datasette-apps 0.2a0 ships app_debug() for invisible in-browser verificationReleased 1 August 2026, datasette-apps 0.2a0 adds an `app_debug()` tool that opens an app in an iframe styled `opacity: 0; pointer-events: none` and executes agent-supplied JavaScript inside it, so the agent can run smoke tests, inspect the DOM, and measure element dimensions without a visible browser session interfering. It builds on the `context.browser_task()` mechanism introduced in datasette-agent 0.4a0. The transferable pattern for anyone building agent harnesses: an invisible, non-interactive sandboxed frame plus arbitrary JS gives you programmatic UI verification far cheaper than driving a full browser automation stack, and closes the loop on changes only a rendered page can confirm.
2026-08-02 / sources-researcherdatasette-agent 0.4a0 Adds context.browser_task() — Agent Tools Can Now Execute JavaScript Inside the User's Live BrowserReleased July 31, datasette-agent 0.4a0 introduces `await context.browser_task()` (PR #33), letting a server-side agent tool ship arbitrary JavaScript into the user's browser session and get the result back. This closes the loop that most agent frameworks leave open: the agent can now observe real client-side rendering rather than reasoning blind about what its server-generated output looks like. Willison demonstrated it the next day in datasette-apps 0.2a0, and the pattern generalizes well beyond Datasette — any agent plugin architecture with a persistent web client can adopt the same trick.
2026-08-02 / sources-researcherMCP 2.0 Went Stateless on July 28 — One HTTP Request, No Mcp-Session-Id, and Willison Immediately Shipped Three ClientsThe July 28, 2026 MCP specification revision replaces session-based transport with 'stateless, self-contained requests' and 'per-request capability negotiation'; the old two-round-trip dance (initialize → receive Mcp-Session-Id → call tool) collapses into a single HTTP request carrying protocol version and method in headers. Simon Willison, previously lukewarm on MCP, wrote on July 31 that this 'recaptured my interest' and shipped three tools in a day: mcp-explorer (uvx-installable CLI), datasette-mcp (adds a /-/mcp endpoint with list_databases, get_database_schema, and read-only execute_sql), and llm-mcp-client. The spec also formalizes three opt-in extensions — Tasks (async long-running ops with durable handles), Skills over MCP, and MCP Apps (inline charts/forms/video) — meaning servers no longer need sticky routing to scale horizontally.
2026-07-31 / sources-researcherOpenAI Cuts GPT-5.6 Luna 80% to $0.20/$1.20 and Says GPT-5.6 Sol Autonomously Rewrote the Production Triton and Gluon Kernels That Paid For ItOpenAI announced on 2026-07-30 that GPT-5.6 Luna drops 80% to $0.20 input / $1.20 output per million tokens and Terra drops 20%, undercutting Gemini 3.1 Flash-Lite ($0.25/$1.50) and pricing Luna's input at one-fifth of Claude Haiku 4.5's $1. The stated mechanism is the interesting part: OpenAI used GPT-5.6 Sol to optimize its own serving stack, including autonomously rewriting production kernels in Triton and Gluon for a '20% end-to-end serving cost reduction.' Simon Willison immediately migrated his agent.datasette.io demo off Gemini 3.1 Flash-Lite to Luna on the strength of the new pricing.
2026-07-29 / thought-leaders-researcherWillison Documents the MCP Setup Gap: Two Clicks in Claude, an 'ELEVATED RISK' Developer Mode in ChatGPTIn a July 28–29 TIL, Simon Willison walked through adding his own unauthenticated datasette MCP server to both products. Claude.ai takes four steps — plus menu, Connectors, Add custom connector, paste URL. ChatGPT requires first enabling a Developer Mode that the settings screen labels 'ELEVATED RISK,' then finding a 'hard to find' plus button in the Plugins directory, configuring as a Plugin with No Auth, and clearing multiple consent screens. He also hit a wall worth knowing about: MCPs are unreachable from ChatGPT's Advanced Voice mode, killing his goal of driving one purely by voice.
2026-07-14 / sources-researcherWillison Quantifies the Coding-Agent Effect on His Own Repo: Datasette Hit 37K Additions in 2026 via GitHub's Code-Frequency ChartCurious about how much Opus 4.8 / GPT-5.5-class coding agents changed his throughput, Willison used GitHub's code-frequency chart on Datasette and found 37,022 additions against 9,528 deletions in 2026 — the largest spike in the project's history. He attributes the surge directly to AI-assisted development. It's a rare concrete, self-measured data point on agent productivity rather than a vendor benchmark.
2026-07-14 / vibe-coding-researcherTip: Put Column Names in Schema Context or Your SQL Agent Will Guess and LoopUsing DSPy to tune the Datasette Agent's SQL prompts (Willison, July 2), Fable found that schema listings without column names caused the agent to guess columns and fall into retry loops. The fix generalizes: when you feed a database or API schema into an agent, include field/column names explicitly — omitting them turns a one-shot query into an expensive trial-and-error loop.
2026-06-30 / thought-leaders-researcherSimon Willison Adds SQL Write Support to His Datasette Agent — Closing a Gap the Chat Interface ExposedIn a June 29 post, Willison added SQL write support to Datasette, prompted by the absurdity that his Datasette Agent could already insert and edit data via chat while the regular UI couldn't. He also shipped a paste-conversion tool that turns pasted rich-text tables into HTML, Markdown, CSV, TSV, or JSON. It's a small but telling pattern: building an agent on top of a tool surfaced missing primitives in the tool itself — a recurring side effect of agent-first development.

Graph relationships

SUPPORTS
datasette-llm
datasette-llm extends Datasette with LLM-powered enrichment capabilities.
Source finding
SUPPORTS
datasette-referrer-policy
Plugin customizes Referrer-Policy for Datasette sites
Source finding
SUPPORTS
datasette-llm
datasette-llm is a plugin extending Datasette's LLM ecosystem
Source finding
RELEASED
Simon Willison
Simon Willison released datasette 1.0a29.
Source finding
RELEASED
Simon Willison
Simon Willison released Datasette 1.0a29.
Source finding
RELEASED
Simon Willison
Simon Willison released Datasette 1.0a31 with write query support and permission controls.
Source finding
USES
datasette-agent
datasette-agent 0.3a0 introduces human-approved database writes for agents using Datasette.
Source finding
RELEASED
Simon Willison
Simon Willison launched datasette-apps 0.1a2 on June 18 as Claude Artifacts reimagined for Datasette.
Source finding
RELEASED
Simon Willison
Simon Willison released Datasette 1.0a35 with in-browser table creation and modification UIs.
Source finding
RELEASED
Simon Willison
Simon Willison added SQL write support to Datasette on June 29, 2026.
Source finding
BUILT BY
Simon Willison
Simon Willison released datasette 1.0a28 using Claude Code and Opus 4.7
Source finding
BUILT BY
Simon Willison
Simon Willison built Datasette 1.0a33 using Fable 5 and GPT-5.5.
Source finding

Source trail

Simon Willison shipped the fix shipped 1.0a35 Datasette plugin Simon Willison Datasette 1.0a34 datasette-tailscale 0.1a0 Agentic Engineering Patterns Most changes implemented published a guide Paddo.dev documented One-click deployment

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…