No, it moved. Elena Verna's November 5, 2026 post says feature discovery becomes largely unnecessary when the customer states a goal and the AI figures out which capabilities to invoke, and her own aside is the tell: at Lovable she makes a scene in Slack when the agent does not know about a feature. That is a zero-adoption ticket filed by the one user who reads the agent's manual. Paweł Huryn's harness post the same day measures about 8,200 tokens of instructions and tool manuals in the window before anyone types, two thirds of it one tool's manual, and calls the words the agent reads part of the product. The tool list is the new navigation. In my own agent-maintained library 30 of 51 commits had no author, so I replaced the skill count with fires per skill and acted-on per fire. A capability with no fires in a month comes out of the manual, because every line costs tokens on every call. Engagement falling while value rises is why the price moves to the outcome; my migration reached 91% outcome revenue.
No. It moved into the harness as the tool list and the skill set, the adoption curve came with it, and now every unused capability costs tokens on every call. The metric is fires per tool and acted-on per fire.
The argument, Huryn's 8,200 tokens, and my own library numbers are in Elena Verna Says Feature Discovery Is Ending. It Moved Into the Harness. Count Fires per Tool.. The fire count comes from 43 Skills, 39 Auto-Activate. Kill the Skill Count. Keep the Four You Type..
Last reviewed 2026-11-05 · 1 min read