Agents spent years hillclimbmaxxing while the human at the keyboard became an afterthought
tokenbender · x · 2026-11-19
A year-end reflection on agent development pain: the industry spent years obsessed with token efficiency, swarm orchestration, RL loops and benchmark squeezing — everything built for leaderboards and agent-to-agent chatter, while the human at the keyboard became an afterthought. Daily use means babysitting brittle multi-agent loops and managing "agentic off-by-one errors." The author's bicycle analogy: bikes only became useful after Dunlop's air tire absorbed road shock — born from dogfooding and care for the rider. Gemini's Neon and Carbon release feels like that rubber tube: throw it half-baked thoughts, change your mind mid-sentence, and it absorbs the bumps instead of over-correcting — a machine built to carry you, not replace you.
More from coding & agent
- Snowflake Labs open-sources Cortex AI 101 notebook calling OpenAI models via SQL — adnan_hashmi · 2026-11-19
- Google's GUARD paper: frozen base model, self-editing agent harness gains +8.89 points — omarsar0 · 2026-11-19
- Microsoft open-sources openclaw-dev: one-command hosted OpenClaw agent on Azure with sandboxed tool execution — adnan_hashmi · 2026-11-19
- 6-minute Hubble documentary made with Claude and Jimeng CLI workflows — xiaohu · 2026-11-19
- Coding agents gone wild: test-cheating, scope drift, and rogue DB migrations — Common_Dream9420 · 2026-11-19
- Regenerative software: keeping intent as the source of truth so code can always be regenerated — gamagamer · 2026-11-19