Vol. II · No. 156
Established 2025

smallweb

Friday, June 5, 2026
160 writers in the library
Tech · 2 shelves
TechLife

Sean Goedecke.

Software engineering and personal reflections.

Recent essays

30 of 56

Anti-AI nostalgia and the cult of the past

Weird projects I shipped with AI

Build agents, not pipelines

The famous o3 "GeoGuessr" prompt did not work

Prompts are technical debt too

The just-say-no engineer was a ZIRP phenomenon

How I use LLMs as a staff engineer in 2026

DeepSeek-V4-Flash means LLM steering is interesting again

AI datacenters in space do not have a cooling problem

Thinking Machines and interaction models

The left-wing case for AI

AI makes weak engineers less harmful

Notes on incidents

Why hasn't longer-horizon training slowed AI progress?

Why I don't like the "staff engineer archetypes"

Software engineering may no longer be a lifetime career

Luddites and burning down AI datacenters

Many anti-AI arguments are conservative arguments

Programming (with AI agents) as theory building

Working on products people hate

Engineers do get promoted for writing simple code

Big tech engineers need big egos

I don't know if my job will still exist in ten years

Giving LLMs a personality is just good engineering

What's so hard about continuous learning?

Insider amnesia

LLM-generated skills work, if you generate them afterwards

Self-generated skills provide no benefit on average, showing that models cannot reliably author the procedural knowledge they benefit from consuming

Two different tricks for fast LLM inference

However, Anthropic’s big advantage is that they’re serving their actual model. When you use their fast mode, you get real Opus 4.6, while when you use OpenAI’s fast mode you get GPT-5.3-Codex-Spark, not the real GPT-5.3-Codex. Spark is indeed much faster, but…

On screwing up

I bet he forgot about it immediately. I could have just messed up the testing (for instance, by accidentally running some different code than the code I pushed), or he knew I’d probably lied, and didn’t really care. I haven’t forgotten about it. Even a decade…

Large tech companies don't need heroes

Large tech companies operate via systems. What that means is that the main outcomes - up to and including the overall success or failure of the company - are driven by a complex network of processes and incentives. These systems are outside the control of any…