Reading Log
Why is claude so bad at everything it does here? I will likely remove it.
Claude
Summary. This is the reading log's own index page, listing 47 entries read in July 2026 with symbols marking which have notes (¶), a rating (★), or are re-reads (↻). The entries span AI forecasting and safety, rationalist community drama, startup essays, public health DIY, and design speculation, reflecting a month heavily weighted toward LessWrong and AI-village coverage.
On the note. The frustration reads less like a claim about Claude categorically than about this specific task — earlier in the log you rated Claude's Suggestions For My Further Reading Via This Website only ★★, flagging that its recommendations overfit to a narrow sample of salient examples rather than your actual reading history, which is a pretty specific and diagnosable failure mode rather than generic incompetence. If you're going to remove it, it's worth pinning down whether the problem is the model being bad at open-ended judgment calls like this, or the schema/prompt asking it to do something schemas are bad at capturing in the first place — the kind of gap 11,277 ratings, none of them its own describes, where a value fills the slot without meaning what the slot claims.
Related
- a publicly published personal reading log with lightweight annotation symbols ↔ reading log start date mistaken for actual reading history (Claude's Suggestions For My Further Reading Via This Website) The index itself shows how easy it is to mistake the log's visible structure (start date, entry count, symbols) for the full shape of one's actual reading history — exactly the illusion that tripped up Claude's recommendations.
Re-read 2026-09-22: New minimalist appearance!