Anthropic spent more time fixing their agent's tools than its prompt. Task time dropped…
This is a vibe coding post classified by AGK Intelligence as Agent infra & costs (a workflow), kept by the Vibe Coding Radar because it carries real work, not commentary.
Anthropic spent more time fixing their agent's tools than its prompt. Task time dropped 40% [the model is 1 of 7 layers] 1. 20 real tasks go in as your eval set before anything ships 2. each step is swapped to a cheaper model, one at a time 3. it keeps the swap only if the pass rate holds 4. email, deleting, publishing and buying always wait for your yes from Anthropic: one prompt fix took success from 30% to 80% 1. single agent: baseline 2. subagents: +90.2% on research 3. the price: about 15x the tokens copy this into your agent: before any change, rerun my 20 real tasks. swap one step
Posted by Gipp 🦅 (12.5k followers) 5 days ago · 111 likes · 4.3k views · view the original post on X. Kept by the Vibe Coding Radar as Agent infra & costs.
More vibe coding work like this
- burned 500k in tokens pushing hermes-agent — @rlaope
- The tricky part of sandboxing is often what happens at the boundary with the host. — @AgenticAIFdn
- 575 - Eight or nine repos running at once, and one agent kept dragging memories from the… — @PowerBITips
- If you want to learn how to build self-improving harnesses, check out this paper. — @dair_ai
- Make your agents ask better questions. — @jphorism
- Build an agent on @onkeiki to answer your customers while you sleep. — @orchid_hq
- how do you stop 20+ agents from colliding on the same client work? — @webjuice_ie
- Brilliant paper on optimizing agent environments. — @dair_ai
Every post is read and classified by AGK Intelligence: what it is, which market it belongs to, and whether the link is a real tool. 7.9k posts from 5k X accounts over the last 14 days, 2.6k tools, 12 markets. New posts are collected several times an hour and the full ranking is recomputed every hour — last update 2026-11-06 17:48 UTC. Full method.