Roadmap

Now

• Depth before breadth. Filling models, agents and concepts properly before opening the long tail. A hub with forty complete entries is worth more than four hundred thin ones — and on a domain assessed as a whole, the thin ones actively cost the good ones.
• Comparison pages for the pairs with genuine demand, written one at a time.
• The core learn tracks: LLM foundations, retrieval, and agents.

Next

• Benchmark tables on model pages, sourced from published evaluations with the evaluation date attached.
• Pricing history, so a model page can show what a price was as well as what it is.
• A public API over the entity graph. The data is already structured for it; what is missing is a rate-limiting and terms story we are happy with.
• Change alerts — tell us which entries you care about and hear when a verified fact on one of them changes.

Later

• Deeper graph queries: "coding agents that run on open-weight models and can be self-hosted" is two hops through data we already hold.
• Non-English coverage, starting with the concepts glossary, which is the part that translates without going stale.

Deliberately not doing

• A generated comparison matrix. See the editorial policy.
• User reviews and star ratings. We cannot verify them, and unverifiable ratings would undermine the one thing this site is for.
• A news firehose. News here exists to date changes to entities, not to compete with newsletters.
• Paid placement, in any form.