03 · a weekend
Now you can read the scoreboard: which models exist, what they cost, and what they can actually be shown to do. The half-life on this knowledge is short, so the point is not memorising the facts. The point is building the habit of checking.
Most confident statements about AI capability come from people whose picture is eighteen months old, which here is a very long time. The instruments and the habit: which models exist, what they cost, what they can be shown to do. Sources over opinions.
Before opinions, find the tables. Epoch gives you training runs; LM Arena gives you messy preference signals; together they keep you from confidently discussing a model that has already been lapped twice.
Compute, parameters, dataset size and estimated cost for essentially every significant model, with methodology published. Half an hour with the filters gives you a better map of the field than most people writing about it have.
Also in The risk argument, Keep current. Ticking it here marks it there.
30 min epoch.ai · free
Blind pairwise human preference at scale, rendered as Elo. Known weaknesses: it rewards style and formatting, it is gameable, and preference is not capability. One instrument, not the only one.
Also in Keep current. Ticking it here marks it there.
10 min lmarena.ai · free
A contamination-resistant leaderboard with objective tasks that refresh over time and show cost beside score. Not magic, just a cleaner instrument for a field where the test set keeps getting eaten by the training set like a suspiciously convenient snack.
Also in Evals & benchmarks. Ticking it here marks it there.
15 min livebench.ai · free
Ready with the full reading map.
Aggregate usage diagnostics are stored; your question and answer text are not.
The useful skill is not memorising this month's winner. It is knowing where the frontier gets reported, how stale the number is, and when the impressive demo is mostly fog machine with invoice attached.
Quality against price against latency across every major model and provider, updated continuously. The page that turns abstract capability talk into a procurement decision, and where the cost collapse at fixed capability becomes impossible to miss.
Also in Keep current. Ticking it here marks it there.
15 min artificialanalysis.ai · free
The field's annual census: capability, investment, adoption, cost, policy and public opinion, all sourced. Skim the top takeaways on release, then use it year-round as the reference whenever someone quotes a number at you. The adoption-versus-measured-value gap in the economy chapter is the most interesting figure in it and the one most often skipped.
Also in Start here, Economics & infrastructure, Applications, Keep current. Ticking it here marks it there.
20 min – 2 hr hai.stanford.edu · free
The official running log for ChatGPT model changes, retirements, routing and behavior updates. It is not a benchmark, but it tells you what users are actually being moved onto, which is often the part missing from lab-chart arguments.
Also in Keep current. Ticking it here marks it there.
10 min help.openai.com · free
Anthropic's collected model and system cards for Claude releases. Read them when a model launches, mostly for the deltas: capability claims, safety evaluations, refusals, tool use and where the company chose to draw the box around the product.
Also in Keep current. Ticking it here marks it there.
15 min platform.claude.com · free
A downloadable weight file can be genuinely useful without making the whole system transparent. These two instruments separate access, documentation and reproducibility, which is less exciting than shouting open and considerably more informative.
Scores major foundation-model developers against one hundred disclosure indicators spanning data, compute, model behavior and downstream use. The 2025 edition makes the useful distinction between releasing weights and documenting enough of the system to support real scrutiny. Treat the scores as a structured disclosure audit, not a universal quality ranking.
30 min crfm.stanford.edu · free
Defines open-source AI around the freedoms to use, study, modify and share, then specifies the data information, code and parameters needed to exercise them. Its sharpest contribution is refusing to call weights alone an open-source system. Read it before using open, open-weight and source-available as interchangeable compliments.
15 min opensource.org · free