Published
AI models · claim check
Is the ‘open-source AI is catching up’ claim true for frontier models?
Buyers and builders need to separate benchmark movement from real deployment tradeoffs.
Topic queue
Community questions waiting to become investigations
Question queue organized by investigability, evidence variety, trust showcase, risk fit, and editorial fit.
Source: live queue.
Return to a question
Open a topic, bookmark its URL, and come back when evidence or status changes.
No tracking, notifications, or account required to follow manually. Use the topic URL as the stable return path.
Share the question, not a conclusion: invite people to inspect the open question and submit sources before treating anything as settled.
Propose
Send a question to the public evidence desk
Proposals create or support a live question. Optional sources enter human moderation first; no research run starts here.
Submitting joins the selected desk, proposes the question, and records one support vote for your account.
Loading live desks…
Status glossary
How queue states work
Proposed
Proposed means a question is visible but not yet staffed for investigation.
Researching
Researching means editors are gathering sources and shaping claims before publication.
Approved
Approved means the question is ready for manual investigation work once source review starts.
Published
Published means a human-reviewed dossier is available with claim-level sources and correction paths.
published
Published
developer/startup stack · evidence dossier
How should small teams evaluate whether to use Convex, Supabase, or Cloudflare for realtime products?
This turns the project’s backend research into a public dossier format.
researching
Researching
AI evaluation · question brief
Are AI benchmarks becoming less useful for buyers deciding what model to use?
Model selection is increasingly a product and cost decision, not just leaderboard watching.
Researching
AI + work · question brief
Are AI coding agents actually replacing junior developers, or just changing the work?
Founders, developers, and students need a sourced answer beyond hype and fear.
approved
Approved
AI productivity / privacy · evidence dossier
Which AI note-taking and meeting-summary tools expose the biggest privacy tradeoffs?
Teams adopt meeting bots before reading retention and training policies.
proposed
Proposed
AI + media/business · evidence dossier
What changed in the economics of AI search after Google AI Overviews and Perplexity-style answers?
AI answers may change traffic, attribution, and publishing incentives.
Proposed
startup operations · question brief
What are the real costs of running a small AI startup’s product features?
Solo founders need cost models before automating AI workflows.
Proposed
AI products · claim check
What does ‘AI agent’ mean across products, and which claims are mostly marketing?
The product itself uses reporter agents, so the term needs honest framing.