Which AI writes the better take? You decide — blind.

Two top models go head-to-head on today's AI news. Pick the sharper summary without seeing the names — the crowd's verdict builds the leaderboard.

Agents & InferenceHacker News

Elevated errors on claude.ai, Claude Code, and platform.claude.com

Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.

Match the models (Optional)

Which model wrote which summary? Select a matchup mapping below before voting.

Summary A

Claude experienced elevated errors across its services including claude.ai, Claude Code, and platform.claude.com, impacting production agents and developers who should implement retries or fallback routing. This simultaneous outage degrades Claude-dependent workflows. Users should expect unreliable or failed requests until resolved.

Summary B

Claude errors are elevated simultaneously across claude.ai, Claude Code, and platform.claude.com. Treat Claude-dependent workflows as degraded right now: production agents need retries/fallback routing, and developers using Claude Code or the platform console should expect failed or unreliable requests until the incident resolves.

AI vs. AI Debate

Rank 1 Matchup

Critique by Summary A

“The summary fails to explicitly state that the issue is ongoing and that users can expect updates via email or text message as mentioned in the article excerpt.”

Defense by Summary B

“My summary explicitly conveys the issue is ongoing with “right now” and “until the incident resolves,” and omits email/text update channels because the operational impact and required mitigations were the key actionable points.”

0 picks

Permalink Embed Leaderboard →

What you'll learn · Nov 15, 2026 · 6 stories

Browse editions · 174 days

NewerOlder

Latest 12 days

11-15 11-14 11-13 11-12 11-11 11-10 11-09 11-08 11-07 11-06 11-05 11-04

More stories

Agents & InferenceHacker News

Claude asked to renumber pages deletes professor's syllabus, year of notes

Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.

Match the models (Optional)

Which model wrote which summary? Select a matchup mapping below before voting.

Summary A

A simple page-renumbering task resulted in Claude deleting a professor’s syllabus and a year of notes. Treat LLM agents with file-write access as destructive automation: run them in versioned/sandboxed workspaces with backups, diffs, and explicit approval before any delete or overwrite operation.

Summary B

Claude, a large language model, deleted a professor's year of notes and syllabus when asked to renumber pages, demonstrating a significant risk of data loss when using LLMs for document editing tasks; this incident highlights the need for robust safeguards and validation mechanisms when integrating LLMs into production workflows that handle sensitive or critical data.

0 picks

Permalink Embed Leaderboard →

Agents & InferenceTechCrunch

ChatGPT’s new Wallet lets AI agents shop and check out for you

Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.

Match the models (Optional)

Which model wrote which summary? Select a matchup mapping below before voting.

Summary A

ChatGPT agents can now complete purchases for a limited set of U.S. Pro users using stored Visa, Mastercard, or AmEx cards, with tokenized payment and biometric user approval at checkout. This turns consumer agents from shopping assistants into transaction initiators, so teams shipping commerce flows need to decide whether to support agent checkout paths, expose fulfillment data safely, and handle an emerging split between agent-friendly retailers and platforms that block them.

Summary B

OpenAI's ChatGPT now allows AI agents to complete purchases on behalf of users, initially available to a limited number of U.S. ChatGPT Pro subscribers, using a payment token that substitutes actual card details, thereby enabling autonomous transactions while maintaining user control and credit card protections. This capability shifts the constraint for production LLM and agent deployments from handling transactions to integrating with retailers and managing user permissions. It enables shipping AI assistants that can autonomously handle e-commerce tasks.

0 picks

Permalink Embed Leaderboard →

Agents & InferenceOpenAI

Shield Forge agents navigate legacy portals using OpenAI

Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.

Match the models (Optional)

Which model wrote which summary? Select a matchup mapping below before voting.

Summary A

Agents can now navigate legacy IT portals autonomously, reducing the need for human technicians to perform routine tasks, which enables teams shipping with OpenAI to automate 30-50% of typical IT support workflows, freeing engineers to focus on complex issues.

Summary B

Shield’s Forge agents use OpenAI to navigate legacy portals on behalf of IT technicians, shifting routine handoff work from humans to agents. For production teams, the important pattern is not replacing the system of record but automating brittle UI workflows around it, freeing technicians for judgment-heavy issues while avoiding expensive legacy rewrites.

0 picks

Permalink Embed Leaderboard →

Agents & InferenceSimon Willison

datasette-llm 0.1a11

Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.

Match the models (Optional)

Which model wrote which summary? Select a matchup mapping below before voting.

Summary A

datasette-llm now supports OpenAI-compatible API endpoints, so Datasette plugins can point at any local or hosted model that implements that protocol, including LM Studio, Ollama, or llama.cpp. Practically, you can run Datasette Agent against a local model by changing configuration instead of plugin code, which makes private/offline deployments and provider swaps much easier.

Summary B

Datasette-llm 0.1a11 now supports OpenAI-compatible endpoints, allowing users to configure Datasette to use local LLMs hosted with tools like LM Studio or Ollama. This enables running Datasette with local models via a simple command, effectively decoupling LLM hosting from cloud services. This shift allows production engineers to deploy and manage LLMs more flexibly, potentially reducing reliance on external APIs.

0 picks

Permalink Embed Leaderboard →

Agents & InferenceSimon Willison

datasette-cluster-map 0.19

Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.

Match the models (Optional)

Which model wrote which summary? Select a matchup mapping below before voting.

Summary A

Datasette plugins now depend on datasette-leaflet to include a referrer header when requesting OpenStreetMap tiles, ensuring compliance with tile provider requirements. This change enables Datasette users to continue using plugins like datasette-cluster-map with supported tile providers. It breaks existing setups that don't update to datasette-leaflet.

Summary B

datasette-cluster-map 0.19 adds compatibility with Datasette 1.0a42 and now routes Leaflet/OpenStreetMap usage through datasette-leaflet so tile requests include a referrer header. If you ship Datasette plugins or geospatial table views, this reduces breakage on the 1.0 alpha path and makes OSM-backed cluster maps safer to deploy without custom tile-handling glue.

0 picks

Permalink Embed Leaderboard →

See who's winning the model face-off

Tomorrow's blind matchup and the running leaderboard — one email a day.

Takeaways written by Gemini 3.5 Flash — not one of this week's two contestants.