Counterparty

Essay · March 1, 2026 · 14:17

Learning to Code is a Waste of Time Now..

The Signal

After hearing Boris Cherny say Claude Code writes every line of Claude Code, then asking five software engineers how they now work, Threadguy argues that AI coding crossed from a promised future into present reality around the release of Claude Opus 4.6. The engineers he recounts — including a quant and a LinkedIn engineer — describe moving within weeks from writing most code by hand to producing it entirely through AI. His conclusion is that code is becoming a commodity, ideas and distribution are the new bottlenecks, and the only honest way to judge the market implications is to use the tools yourself.

Key Takeaways

  • 01

    The change took weeks

    A quant friend says he went from writing 60–70% of his code by hand to writing none; a LinkedIn engineer describes a similar move from 80–90% by hand to zero. Threadguy hears the same pattern across his weekend conversations and dates the break to the arrival of Claude Opus 4.6.

  • 02

    Code moves into English

    If AI produces every line, Threadguy argues, knowing syntax is no longer the central bottleneck. He still says people need to understand how systems work, but puts the scarce inputs elsewhere: ideas, judgment, and distribution.

  • 03

    Use the tool, then take a side

    Threadguy rejects forming an AI market view from a distance. His test is to try the tools, decide how good and consequential they are, and treat either informed conclusion — bullish or skeptical — as potential edge.

On the Record

100% of code is being written by AI. I mean, that is like the most insane, ridiculous thing I've ever heard. It's always been in six months, in six months, in six months, in six months. But, brother, I think it's here. I think we just woke up one day and it's here.

The only way to do that is you got to go try it. You have to just try and make your own decision. If you try it and you decide it's not that good, that's edge. That's edge. Everyone is moonboying. It's not that good. That is edge. If you try it and you decide this is the greatest thing of all time, that is edge. That's edge.

The Breakdown

The tweet that broke the six-month loop

Threadguy starts on December 27, 2025, when Boris Cherny replied that Claude Code wrote 100% of its own code. After years of hearing transformative AI was six months away, he treats that answer as a line crossed rather than another forecast. Moltbook's viral, fully vibe-coded experiment, then the February 5 releases of Claude Opus 4.6 and an OpenAI coding model, made the change feel visible; OpenClaw was his own entry point.

Five engineers, one abrupt reset

To see whether the shift was mostly hype, he asks five software engineers how much code AI is writing. His quant friend says he had still been writing 60–70% by hand a month before but had written none for weeks; a LinkedIn engineer describes a move from 80–90% by hand to thousands of AI-produced lines without a manual keystroke. A third engineer gives the same post-4.6 answer: zero by hand, 100% AI.

Why adoption stayed quiet

His quant friend describes a six-to-12-month social contract among engineers not to advertise AI use: fear for their jobs, taboo, pride in writing code, and legal or security concerns. Threadguy says the striking part is that these workers use the same public Opus and Codex models available to everyone outside the frontier labs. In his framing, code is now a commodity written through English; software knowledge still matters, but ideas, judgment, and distribution become the bottlenecks.

From workflow shift to market position

From there he turns the workflow change into a market thesis: he says it is too large and too recent to be priced in, and predicts a surge of software creation, a new roaring '20s, and sustained compute demand. He contrasts the engineers' reports with a skeptical Bloomberg analyst who had not tried OpenClaw, and guesses that only 500,000 to one million people worldwide have messaged it. His practical test is simpler than the forecast — use the tools, decide how capable and consequential they are, and treat either informed conclusion as edge; by the end, he says his zero stock exposure is about to change.

Distilled from the episode transcript · Counterparty Recap Desk

More from the desk

View all