Malleable agents: the shift isn't more reasoning, it's systems that expand what they can do
dfinke · x · 2026-09-03
The author offers a mental framework for classifying agents:
- Known path + known tools = automation
- Unknown path + known tools = fixed agent
- Unknown path + potentially unknown capabilities = malleable agent
The key claim: what matters isn't just stronger reasoning, but systems that can keep exposing capabilities unknown at design time. To find where malleability matters, look at work that keeps revealing capabilities the system didn't have when designed—not merely whether an agent can solve a given problem.
Related event: A simple framework: automation, fixed agents, and malleable agents(2 posts)→
More from coding & agent
- How I Wrangle AI Agents — a practitioner writeup — dfinke · 2026-09-04
- One founder turned his endless idea backlog into an AI-built public archive site — Delahuntagram · 2026-09-04
- Teacher With 16GB VRAM Hits a Wall: Local LLMs Keep Failing at MCP Tool Use — whakahere · 2026-09-03
- ComfyUI-Minimax-H3-Promptor v1.5 Adds Shot-Level Refine, Autocomplete, Local GGUF — Narrow-Particular202 · 2026-09-03
- OpenClaw 2.0 adds per-user OAuth for MCP servers, enabling multiplayer shared bots — steipete · 2026-09-03
- PRs NOT Welcome: Vercel, tldraw and Flue replace community PRs with agent software factories — rseroter · 2026-09-03