From the Trenches
Subscribe
Sign in
Home
Archive
About
Latest
Top
Discussions
The Bias Lives in the Weights
Corporate bias in LLMs is architectural. It enters through training, surfaces at inference, and survives open-sourcing.
May 19
•
Denis Stetskov
15
4
4
Silicon Valley Eats the War
Meta signed a $6B fiber deal. Same month, a Ukrainian drone factory halted orders. Same fiber. Same factories. Different buyers. The math doesn't work.
May 12
•
Denis Stetskov
20
2
8
AI Is a Mirror of Our Engineering Culture
CMU tracked 807 repos after Cursor adoption. Complexity up 41%. Warnings up 30%. Copilot output now trains the next model. The feedback loop is already…
May 5
•
Denis Stetskov
59
13
17
April 2026
I Was Wrong About Anthropic
Six months ago I called Anthropic "responsible AI done right." Their models got worse, their CPO burned Figma, and Claude picks targets in Iran.
Apr 28
•
Denis Stetskov
51
8
13
The West Forgot How to Make Things. Now It’s Forgetting How to Code
The defense industry lost the ability to make weapons when crisis hit. The same pattern is eroding software engineering skills. The timelines are…
Apr 21
•
Denis Stetskov
312
64
57
Everyone Wants a Better Team. Nobody Wants to Do Anything About It.
Same meeting. Scorecard says zero problems. Out loud, the same engineers describe a dozen. The gap between what people say and what they write
Apr 14
•
Denis Stetskov
21
26
3
The Human Cost of 10x: How AI Is Physically Breaking Senior Engineers
AI tools increased code review volume by 98% but your brain still runs at 10 bits per second. The physical toll on senior engineers is measurable.
Apr 7
•
Denis Stetskov
244
52
47
The Snake That Ate Itself: What Claude Code’s Source Revealed About AI Engineering Culture
Anthropic claimed 100% of Claude Code is AI-written. A source leak exposed a 3,167-line function, regex sentiment analysis, and 250K wasted API calls…
Apr 1
•
Denis Stetskov
890
162
172
March 2026
Your CLAUDE.md Is a Wish List, Not a Contract
AI coding agents follow fewer than 30% of instructions perfectly. Real compliance data from AGENTIF, METR, and thousands of supervised sessions with…
Mar 30
•
Denis Stetskov
30
7
8
The Autonomy Illusion
A benchmark gave 12 AI models a food truck. 8 went bankrupt. Every model that borrowed money failed. Here's what that means for your codebase. And the…
Mar 23
•
Denis Stetskov
15
4
AI’s Announcement Problem
Conference stages say AI replaces engineers. Production data says otherwise. I have both numbers.
Mar 16
•
Denis Stetskov
22
4
Your Brain on Autopilot: The Cost of AI Thinking for You
Eighty-three percent of ChatGPT users couldn’t recall key points from essays they had written minutes earlier.
Mar 9
•
Denis Stetskov
30
10
6
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts
