---
schema: 1
type: video
title: >-
  Coding Agents Don't Scale Themselves. Neither Do Your Teams. — Patrick Debois,
  Tessl
description: >-
  In 2009 people told Patrick Debois that continuous delivery was crazy. He
  hears the same thing now about the dark factory, and reads it the same way:
  not that the technology cannot work, but that the organization is not set up
  for it yet. His starting assumption is that harnesses and loops will
  commoditize, possibly into a service a frontier lab just sells you, so none of
  that will be anyone's differentiator. What actually changes is the team, the
  platform and the organization around them. Developers pushed back on the
  conductor framing, telling him they did not sign up to write better prompts.
  What brought them back was tooling. Once the team started building harnesses
  for the agent, a genuinely technical path reopened, and the loudest skeptics
  turned out to be the right people to hand context authoring to, precisely
  because they were angry about the output. The shift he pushes is to stop
  fixing the code the agent produced and improve the system instead. Retros stop
  being about the code and become about where the agent hit the same wall
  repeatedly. Planning splits into work scoped tightly enough to hand off and
  work that still needs a conversation. He tracks two numbers: how many human
  touches it takes to get the right result, which should fall, and how much of
  each fix is shared, because one improvement to a common harness lands for
  everybody rather than making a single person 10x. Above the team it is paved
  roads and a named owner, not a thousand flowers blooming. Speaker info: -
  https://x.com/patrickdebois - https://www.linkedin.com/in/patrickdebois/ -
  https://jedi.be Timestamps: 0:00 - It will not work here, and what that
  signals 2:46 - Developers who did not sign up for prompting 5:20 - Stop fixing
  the code, improve the system 7:04 - Retros, planning, and the downstream
  squeeze 9:38 - Two metrics: human touches and reuse 10:29 - The platform
  team's new problems 12:12 - Sprawl, paved roads, and making spend visible
  14:43 - Enabling the organization without a champions program 15:36 - Hiring
  when the job titles mean nothing 18:07 - Optimize the spend instead of capping
  it 20:38 - A dim factory, and knowledge as the moat --- Watch along with the
  corrected transcript, summary, timestamps and resources on this talk's
  official AIE page:
  https://ai.engineer/talks/zCJtYuqwm7E-youtube-3e8e342eb2b6ce07f253 Join our
  upcoming AI Engineer conferences (see https://ai.engineer landing page): -
  AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and
  engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference
  returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS
  2026 in Sydney ---
canonical: 'https://liquidgold.cc/watch/zCJtYuqwm7E'
markdown: 'https://liquidgold.cc/watch/zCJtYuqwm7E.md'
llms: 'https://liquidgold.cc/watch/zCJtYuqwm7E/llms.txt'
parent: 'https://liquidgold.cc/channel/aidotengineer.md'
channel: aidotengineer
updated: '2026-10-24T04:56:02.177Z'
language: null
children: []
videoId: zCJtYuqwm7E
durationSeconds: 1326
publishedAt: '2026-08-22T16:00:06.000Z'
searchChunks:
  - ordinal: 0
    text: null
    startSeconds: 23
    endSeconds: 170
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=23'
    heldReason: rights
  - ordinal: 1
    text: null
    startSeconds: 170
    endSeconds: 182
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=170'
    heldReason: rights
  - ordinal: 2
    text: null
    startSeconds: 182
    endSeconds: 327
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=182'
    heldReason: rights
  - ordinal: 3
    text: null
    startSeconds: 182
    endSeconds: 327
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=182'
    heldReason: rights
  - ordinal: 4
    text: null
    startSeconds: 327
    endSeconds: 338
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=327'
    heldReason: rights
  - ordinal: 5
    text: null
    startSeconds: 338
    endSeconds: 429
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=338'
    heldReason: rights
  - ordinal: 6
    text: null
    startSeconds: 429
    endSeconds: 439
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=429'
    heldReason: rights
  - ordinal: 7
    text: null
    startSeconds: 439
    endSeconds: 578
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=439'
    heldReason: rights
  - ordinal: 8
    text: null
    startSeconds: 439
    endSeconds: 578
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=439'
    heldReason: rights
  - ordinal: 9
    text: null
    startSeconds: 439
    endSeconds: 578
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=439'
    heldReason: rights
  - ordinal: 10
    text: null
    startSeconds: 578
    endSeconds: 588
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=578'
    heldReason: rights
  - ordinal: 11
    text: null
    startSeconds: 588
    endSeconds: 632
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=588'
    heldReason: rights
  - ordinal: 12
    text: null
    startSeconds: 643
    endSeconds: 736
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=643'
    heldReason: rights
  - ordinal: 13
    text: null
    startSeconds: 643
    endSeconds: 736
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=643'
    heldReason: rights
  - ordinal: 14
    text: null
    startSeconds: 736
    endSeconds: 746
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=736'
    heldReason: rights
  - ordinal: 15
    text: null
    startSeconds: 746
    endSeconds: 885
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=746'
    heldReason: rights
  - ordinal: 16
    text: null
    startSeconds: 885
    endSeconds: 898
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=885'
    heldReason: rights
  - ordinal: 17
    text: null
    startSeconds: 940
    endSeconds: 952
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=940'
    heldReason: rights
  - ordinal: 18
    text: null
    startSeconds: 952
    endSeconds: 1088
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=952'
    heldReason: rights
  - ordinal: 19
    text: null
    startSeconds: 1088
    endSeconds: 1100
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=1088'
    heldReason: rights
  - ordinal: 20
    text: null
    startSeconds: 1100
    endSeconds: 1248
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=1100'
    heldReason: rights
  - ordinal: 21
    text: null
    startSeconds: 1100
    endSeconds: 1248
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=1100'
    heldReason: rights
  - ordinal: 22
    text: null
    startSeconds: 1248
    endSeconds: 1261
    watch: 'https://liquidgold.cc/watch/zCJtYuqwm7E?t=1248'
    heldReason: rights
---


# Coding Agents Don't Scale Themselves. Neither Do Your Teams. — Patrick Debois, Tessl

In 2009 people told Patrick Debois that continuous delivery was crazy. He hears the same thing now about the dark factory, and reads it the same way: not that the technology cannot work, but that the organization is not set up for it yet. His starting assumption is that harnesses and loops will commoditize, possibly into a service a frontier lab just sells you, so none of that will be anyone's differentiator. What actually changes is the team, the platform and the organization around them. Developers pushed back on the conductor framing, telling him they did not sign up to write better prompts. What brought them back was tooling. Once the team started building harnesses for the agent, a genuinely technical path reopened, and the loudest skeptics turned out to be the right people to hand context authoring to, precisely because they were angry about the output. The shift he pushes is to stop fixing the code the agent produced and improve the system instead. Retros stop being about the code and become about where the agent hit the same wall repeatedly. Planning splits into work scoped tightly enough to hand off and work that still needs a conversation. He tracks two numbers: how many human touches it takes to get the right result, which should fall, and how much of each fix is shared, because one improvement to a common harness lands for everybody rather than making a single person 10x. Above the team it is paved roads and a named owner, not a thousand flowers blooming. Speaker info: - https://x.com/patrickdebois - https://www.linkedin.com/in/patrickdebois/ - https://jedi.be Timestamps: 0:00 - It will not work here, and what that signals 2:46 - Developers who did not sign up for prompting 5:20 - Stop fixing the code, improve the system 7:04 - Retros, planning, and the downstream squeeze 9:38 - Two metrics: human touches and reuse 10:29 - The platform team's new problems 12:12 - Sprawl, paved roads, and making spend visible 14:43 - Enabling the organization without a champions program 15:36 - Hiring when the job titles mean nothing 18:07 - Optimize the spend instead of capping it 20:38 - A dim factory, and knowledge as the moat --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/zCJtYuqwm7E-youtube-3e8e342eb2b6ce07f253 Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney ---

Canonical: https://liquidgold.cc/watch/zCJtYuqwm7E · Channel: [AI Engineer](https://liquidgold.cc/channel/aidotengineer.md)

> Search chunk text is held until public-source provenance is verified.

## Document chunks

> Missing: no stored document chunks are available.
