Uploaded October 2026 | Updated November 2026, 1 day ago
AI writes the code now. So do we still need programmers? I hand a request from accounting to a coding agent, and a few minutes later the report is done. Except it isn't.
Picture the work as a tree: the code is one leaf at the very bottom. Over the last few years, the line of what you can hand to AI has climbed that tree tier by tier, from autocomplete to code from a sentence to agents that write, test and fix in a loop. Where does that line sit today, where did you step off, and what's actually left for programmers? More than you'd think. Just not the same as before.
This is also the first episode I'm putting out as a podcast, so it works with or without the visuals. I did try to improve those though. :)
After wrecking my brain I settled for Partially Uploaded for its name.
Chapters
0:00 The report is done
2:16 The whole tree
4:44 Tier 0: everything by hand
5:35 Tier 1: autocomplete
6:25 Tier 2: reusable blocks
6:53 Tier 3: code from a sentence
8:49 Tier 4: the loop closes
11:16 The messy middle
12:54 Today: the jumps you couldn't see
15:02 Natalie's report isn't done
18:24 Branches that didn't exist three years ago
20:57 Who operates this?
23:56 The top keeps growing
24:56 Load your last save
Sources
Hallucinated package with 30,000+ downloads (Bar Lanyado, Lasso Security): simonwillison.net/2024/Apr/1/diving-deeper-into-ai-package-hallucinations
Stack Overflow Developer Survey 2025, AI: survey.stackoverflow.co/2025/ai
METR, experienced open-source developers with early-2025 AI tools: metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study
GitClear, moved and restructured code: gitclear.com/the_ai_code_quality_maintainability_gap
METR time horizons: metr.org/time-horizons
Bitkom, AI in German companies (September 2026): bitkom.org/Presse/Presseinformation/Erstmals-nutzt-Mehrheit-Unternehmen-KI
Kent Beck, "90% of My Skills Are Now Worth $0": newsletter.kentbeck.com/p/90-of-my-skills-are-now-worth-0
TypeSafe AI's Jev: letsdatascience.com/news/typesafe-ai-launches-jev-decision-model-889a38c0
W. S. Jevons, The Coal Question (1865)
Natalie's report is done. The job isn't.
#AIcoding #Programming #AIagents
AI writes the code now. So do we still need programmers? I hand a request from accounting to a coding agent, and a few minutes later the report is done. Except it isn't.
Picture the work as a tree: the code is one leaf at the very bottom. Over the last few years, the line of what you can hand to AI has climbed that tree tier by tier, from autocomplete to code from a sentence to agents that write, test and fix in a loop. Where does that line sit today, where did you step off, and what's actually left for programmers? More than you'd think. Just not the same as before.
This is also the first episode I'm putting out as a podcast, so it works with or without the visuals. I did try to improve those though. :)
After wrecking my brain I settled for Partially Uploaded for its name.
Chapters
0:00 The report is done
2:16 The whole tree
4:44 Tier 0: everything by hand
5:35 Tier 1: autocomplete
6:25 Tier 2: reusable blocks
6:53 Tier 3: code from a sentence
8:49 Tier 4: the loop closes
11:16 The messy middle
12:54 Today: the jumps you couldn't see
15:02 Natalie's report isn't done
18:24 Branches that didn't exist three years ago
20:57 Who operates this?
23:56 The top keeps growing
24:56 Load your last save
Sources
Hallucinated package with 30,000+ downloads (Bar Lanyado, Lasso Security): simonwillison.net/2024/Apr/1/diving-deeper-into-ai-package-hallucinations
Stack Overflow Developer Survey 2025, AI: survey.stackoverflow.co/2025/ai
METR, experienced open-source developers with early-2025 AI tools: metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study
GitClear, moved and restructured code: gitclear.com/the_ai_code_quality_maintainability_gap
METR time horizons: metr.org/time-horizons
Bitkom, AI in German companies (September 2026): bitkom.org/Presse/Presseinformation/Erstmals-nutzt-Mehrheit-Unternehmen-KI
Kent Beck, "90% of My Skills Are Now Worth $0": newsletter.kentbeck.com/p/90-of-my-skills-are-now-worth-0
TypeSafe AI's Jev: letsdatascience.com/news/typesafe-ai-launches-jev-decision-model-889a38c0
W. S. Jevons, The Coal Question (1865)
Natalie's report is done. The job isn't.
#AIcoding #Programming #AIagents










