Subscribe
Sign in
Home
Notes
Archive
About
Latest
Top
The AI Layoff Story Is Too Convenient
Big tech is not simply replacing workers with bots. It is using AI, higher rates, and a colder labour market to redraw which people count as scarce.
Jun 14
•
Eloi Tay
The Benchmark Is Also Hallucinating
DeepSWE is a serious test of coding agents. The silly bit is pretending any one leaderboard can tell you whether software work is safe.
Jun 11
•
Eloi Tay
The Moat Moved Up the Stack
Open models can catch a leaderboard. That is not the same thing as catching the product.
Jun 9
•
Eloi Tay
We Keep Promoting the Wrong Kind of Leader
We Keep Promoting the Wrong Kind of Leader
Jun 7
•
Eloi Tay
The High Performer Trap
Why the person who saves the day can become the manager who blocks the team.
Jun 4
•
Eloi Tay
GPT-5.5 Is Expensive Only If You Ask the Wrong Question
Price per token is the spreadsheet answer. Cost per useful outcome is the production answer.
Jun 2
•
Eloi Tay
Free Tools to Help You Stretch Your Tokens
RTK cuts terminal noise before it enters context. Caveman cuts the agent’s visible chatter. The useful trick is knowing which tokens each one actually…
Jun 1
•
Eloi Tay
April 2026
AI Is Cheap. Good Ideas Still Aren't.
The useful question is not whether AI saves hours. It is whether it lets teams buy more serious attempts without lowering the bar.
Apr 28
•
Eloi Tay
Your Agent Isn't Tired. It's Drowning in Context
The useful analogy between humans and LLMs is not consciousness. It is that both fall apart when active memory, compression, and management are sloppy.
Apr 25
•
Eloi Tay
Your Org Chart Is Not Your Operating Model
Why healthy software teams keep accountability close to the product, centralize leverage, and design for knowledge resilience
Apr 23
•
Eloi Tay
You Are Probably Rebuilding DSPy Already
Why prompt tables, eval scripts, and model wrappers keep turning into the same architecture.
Apr 22
•
Eloi Tay
The New Senior Engineer Knows What to Trust
In the AI era, the real edge is not writing more code. It is deciding what counts as evidence before the code ships.
Apr 20
•
Eloi Tay
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts