TLDR Dev 2026-06-15
Why AI won’t replace devs 🚫, vibe coders vs SWEs ⚔️, fixing AI frontends 🔨
Slightly reducing the sloppiness of AI generated frontend (2 minute read)
Creating attractive frontends with AI usually leads to “slop”-looking generations. However, asking AI to produce styles as a “Qt app” ends up removing the frontend AI-generated “code smells.”
Formal methods and the future of programming (10 minute read)
Formal methods are techniques that model software as mathematical entities to provide rigorous, proof-based guarantees of correctness rather than relying on standard testing. The rise of agentic coding is now shifting the cost-benefit analysis for these methods, making them an increasingly accessible tool for verifying AI-generated code, whereas before they were too expensive to consistently use.
The only scalable delete in Postgres is DROP TABLE (5 minute read)
Large Postgres DELETEs add work rather than reclaim it. They leave dead tuples, add replication overhead, and don't return disk space, while DROP TABLE and TRUNCATE remove files directly and scale independently of data size. It's best to structure your schema so bulk deletion becomes a DROP or TRUNCATE.
A frontier without an ecosystem is not stable (4 minute read)
The transition to an AI-driven economy requires a cognitive loop where human capital and token capital compound to create unique institutional value. To maintain sovereignty, firms must build private learning loops that encode their domain expertise into agentic systems that improve with each use.
No, everyone is not using AI for everything (4 minute read)
Contrary to the narrative that "everyone is using AI for everything," recent data shows that AI adoption has actually slowed down, with public sentiment turning increasingly skeptical. Real-world usage and survey data show that the US population is roughly split into thirds: one-third actively use AI, one-third use it occasionally, and one-third never use it.
Don't trust large context windows (3 minute read)
LLMs have a reliable "smart zone" that typically degrades into a "dumb zone" once input exceeds approximately 100,000 tokens. Although developers market context windows reaching into the millions, performance studies show that effective attention is only a small fraction of those advertised figures.
Why AI hasn't replaced software engineers, and won't (23 minute read)
AI is unlikely to cause mass layoffs in software engineering because the profession relies on a "decide-execute-deliver" workflow where only the middle execution phase is automated. Recent high-profile job cuts attributed to AI are frequently the result of financial pressures or pandemic-era over-hiring rather than actual technological displacement.
The computer science degree isn't dead (4 minute read)
Despite rising unemployment figures for recent graduates, computer science and engineering remains a top-tier field when measuring overall labor market outcomes and long-term earnings. The current challenge is mostly in a broken hiring pipeline where many entry-level listings serve as "ghost jobs" while actual hiring rates for junior positions have declined.
Vibe Coder vs Software Engineer (10 minute read)
Vibe coding focuses on the rapid generation of prototypes to test ideas, whereas software engineering prioritizes the entire development lifecycle and the complexities of long-term maintenance. The main difference is in measurement, as the former focuses on the speed of a first working version while the latter values the "time to safe merge.”
The most important software engineering news in one daily email
Join 450,000 readers for
one daily email