Articles
June 2026
Doing nothing at work
Performance at tech companies is dominated by outlier events. When I think about the most impactful changes I’ve made, many of them involved a surprisingly trivial amount of work. There are no points for effort in software development. What matters is solving the right problem at the right time.
You Got Faster. Your Company Didn’t.
It’s why Pascal, in that quote at the top, was apologizing: the long letter is cheap for me and expensive for you, the short letter is expensive for me and cheap for you. At the workplace, I usually owe you the short one because there’s one of me and many of you. Compression, editing, and fact-checking are the work.
Old Software Was Fast Because It Had No Choice
The gains did not disappear. We spent them on larger runtimes, deeper dependency graphs, heavier containers, more telemetry, wider safety margins, and platform defaults that make every service look the same. When a higher-level abstraction makes coding more efficient, we do not write less code or build simpler systems. We spend the efficiency on more complexity, more integration, more moving parts.
A return to two-pizza culture
When you have conviction about the customer problem but have genuine uncertainty about whether your approach will work, you start by building a prototype. Then you use it the way a customer would. You break things, you find the gaps your intuition missed, then you share it with a few colleagues, and if there’s excitement around what you’ve built, you write the doc. Having something tangible to click through as you write changes the quality of the document.
May 2026
We Are (Still) Living in the Long Boring
Digital innovations merely refine existing domains — genuine material progress happened between 1870–1970 and requires political action, not tech optimism
Agentic Coding Is a Trap
Agentic coding atrophies the exact skills needed to supervise AI effectively — making you a less capable orchestrator the longer you rely on it
The Code Was Always the Door
Senior developers should embrace an AI shepherd role, leveraging judgment and domain knowledge rather than fearing replacement by automation
Be the Idiot
The only thing more expensive than looking stupid in a meeting is building the wrong thing because everyone was too proud to ask
Appearing Productive in The Workplace
AI enables workers to produce expert-looking output they cannot evaluate, undermining genuine expertise and institutional judgment
When everyone has AI and the company still learns nothing
Individual productivity gains from AI do not automatically become organizational gains — without feedback loops, discoveries stay siloed
On Grindslop
Wealthy tech founders perform extreme self-sacrifice to justify their wealth, mimicking exploitative labor conditions without genuine purpose
April 2026
Go Ahead and Use AI. It Will Only Help Me Dominate You.
Using AI causes cognitive surrender — letting competitors atrophy their skills while you maintain your intellectual edge
The machines are fine. I'm worried about us.
AI agents in academia risk producing publishable results without the deep struggle that builds genuine scientific understanding
A Guide to Context Engineering for LLMs
Effective LLM performance is about strategically managing what enters the context window, not just maximizing its size
Let's talk about LLMs
LLMs address only the accidental difficulties of software development, not the essential ones — specification, design, and testing
Drunk Post: Things I've Learned as a Senior Engineer
The best code is no code at all — and soft skills like documentation and proposals are massively underrated
March 2026
The Illusion of Building
There is still a big gap between making an app in the shape of something vs. making and operating a real software system
Humans and Agents in Software Engineering Loops
Moving from the 'how' loop to creating the harness for the 'how' and still playing the key role in the 'why' loop
Infinite midwit
AI has objective intelligence but lacks subjective intelligence — taste, judgment, and the ability to create work that actually means something
Every minute you aren't running 69 agents, you are falling behind
The AI hype cycle is toxic fear-mongering — the real threat isn't falling behind on adoption, it's rent-seeking becoming obsolete as larger players consolidate those activities
February 2026
Spec-driven development doesn't work if you're too confused to write the spec
Specification-driven development is ineffective when developers lack sufficient clarity to articulate the specifications in the first place
Bring Back Ops Pride (xpost)
Operations engineering deserves respect as a critical discipline, and the industry should stop renaming ops roles to escape the stigmatized term
No Silver Bullet—Essence and Accidents of Software Engineering
There is no single technology or management technique that can deliver an order-of-magnitude improvement in software productivity
Tool Shaped Objects
AI systems create the sensation of productive work rather than actual output, as the market for feeling productive is far larger than the market for being productive
End Game Play
Modern society skips the opening and middlegame phases of complex endeavors to rush toward predetermined endings, creating an illusion of progress while neglecting actual work
January 2026
The AI Ready Software Developer
A series of very rational and clear thinking about how LLMs can be helpful but are very limited
LLMs, Context Windows & Brown M&Ms
How LLMs fail the brown m&m test
AWS and Microsoft are selling much more than cloud services
An extremely important corporate value is never getting blamed
Being creative requires taking risks
People who are able to remain interesting design their lives to avoid [mode collapse] happening to them
Pluralistic: Code is a liability (not an asset)
Code is a liability and the cost is the work to maintain it
Slowing Down AI On Purpose
Use LLMs in small batches of code with the human still in the driver's seat
Agentic AI and The Mythical Agent-Month
Scaling agents up in parallel still hits mathematical limits and is limited by the common knowledge problem