Tuesday, June 23, 2026

Follow builders, not influencers

5 stories · 4 min read

There's a quiet theme running through today's batch: the psychological side of AI tools, not the technical side. How you feel about the work AI does for you turns out to matter as much as whether the work is any good. That tension shows up in two very different ways today.

01

Vercel CEO names the real danger of coding agents

Vercel CEO Guillermo Rauch posted a one-liner that landed hard: "Coding agents will squeeze every ounce of IKEA effect out of you, if you let them." The IKEA effect is the cognitive bias where you value things more because you built them yourself. Rauch is pointing at something most engineers haven't fully admitted yet: when an agent writes your code, you feel less attached to it, less proud of it, and possibly less inclined to defend its quality when someone in a meeting pushes back. ---

Why it matters: This isn't an abstract psychology problem. Engineers who don't feel ownership over their code make worse decisions about it. If your team ships agent-written code nobody feels responsible for, the blame-shifting starts before the first bug lands in production.

Source →

02

The trick that makes AI video agents actually work: write HTML, not video

Product builder Peter Yang shared an insight from a team that spent months trying to build a video agent before discovering the obvious-in-hindsight answer. The problem: AI agents have no visual intelligence. They can't "see" whether a video looks good. The solution the team landed on: generate video as HTML, CSS, and JavaScript instead of as raw footage. Because language models already think in HTML natively, they can control layout, aesthetics, and motion through code rather than through visual judgment they don't have. ---

Why it matters: If you're building anything that asks AI to create visual output, this is the architecture worth copying. The frame "what format does the model actually understand?" is more useful than "what output do I want?" and it applies well beyond video.

Source →

03

Nan Yu on why quality requires an irrational commitment

Nan Yu shared a sharp observation: "Quality is irrational" is the right frame for it. His point is that consistently choosing quality over convenience requires an almost unreasonable level of conviction, plus the self-belief that owning things end-to-end will beat reaching for common frameworks. It's a short post, but it names something real about the gap between teams that ship good products and teams that ship fast products. ---

Why it matters: This cuts directly at the AI-speed moment everyone is in. When agents can generate code, copy, and designs in seconds, the pressure to ship fast gets louder and the case for slowing down to get it right gets harder to make. "Quality is irrational" is a useful reminder that it was always a deliberate choice, not a default.

Source →

04

OpenAI's Codex adds bankable usage resets

Thibault Sottiaux, who works on the Codex App at OpenAI, posted asking how users think about the new ability to bank usage resets in Codex rather than use them immediately. The question pulled 868 replies, which suggests this feature is touching a real nerve about how people meter their AI usage. ---

Why it matters: The fact that this generated nearly 900 replies tells you developers are actively strategizing around AI usage limits. When people start hoarding AI compute the way they hoard vacation days, the pricing model has officially become its own behavioral problem.

Source →

05

Swyx on an insurance startup breaking through in a notoriously frozen market

Swyx mentioned in passing that a startup called Corgi has apparently captured near-total market share among his real estate broker's clients, with the broker saying "just go with corgi, they are covering every single one of my clients right now." He called this level of greenfield penetration "unheard of in the insurance industry." This is thin on specifics, but the observation is worth flagging: insurance is one of the markets AI startups have been circling for years, and a product getting that kind of organic word-of-mouth from brokers suggests someone finally cracked the distribution problem, not just the technology.

Source →

Get this in your inbox

3-5 stories. 3 minutes. Every weekday morning.

Past issues

View all