Marc Lou's $217K month and DHH's case for Rust
Josh Pigford says every sale of a product he has not even shown yet came through Twitter DMs, and he puts the odds of $1M ARR by month end at 97%. Meanwhile DHH had agents rewrite Campfire in three languages and came away arguing the language choice is no longer about you.
Real money
Pigford says a week of sales, all of it through Twitter DMs
Josh Pigford says that one week in, every sale of his unreleased product has come through Twitter. Two days earlier he said he was about 97% certain the thing would be at $1M ARR by end of month, once he flips the switch. He has still not shown the product, and there is no revenue figure in either post.
based on current demand, i'm about 97% certain this thing will be at $1m ARR by end-of-month
Marc Lou reports $217,071 in September at 90% margin, solo
Marc Lou published a September breakdown of $217,071 across 17 products at a 90% margin, run entirely by himself. He says HYROX at $112K inflates the month, but it was also the best ever month for his marketplace TrustMRR at $53K and his SaaS DataFast at $31K. The bottom of the list is small: $143, $97, $9.
I do everything solo, and I'm not going to lie, I'm fucking proud of myself.
Opus 5.5 takes half of all T3 Code prompts, Theo says
Theo says Opus 5.5 is the first model to break 50% of traffic in T3 Code, with literally half of all prompts going to it. dax says Claude is his team's most used model for the first time since the beginning of the year. Theo's own scorecard puts Opus 5.5 at 9/10 on code capability and 8/10 on price, against 3.5/10 on price for GPT-6 Astra.
Preferring OpenAI models for code makes almost no sense.
Stolen playbooks
DHH had agents rebuild Campfire in Elixir, Go and Rust to compare
DHH says he had agents implement and optimize the Campfire web app in Elixir, Go and Rust, and confirms Ruby came out slowest. Guillermo Rauch answered with Vercel's own version of the experiment: Turborepo was ported from Go to Rust, the migration finished, and the return on investment was controversial internally because the human migration costs were substantive. Rauch adds that Rust is unlikely to be the endpoint either.
What's "best for humans" is no longer necessarily "best for business
Founder fights
DHH: a language agents cannot write well is a dying language
Pushed on his Campfire results, DHH said that if frontier agents are not good at writing Elixir or Go, that makes the case worse, not better. He is careful that not every app needs Rust-level performance, but says if you accept that nobody reads every generated line, speed wins the tiebreak. Rauch sided with him and called the old calculus, picking what humans enjoy writing, obsolete.
Any programming language that agents can't write well out of the box is going to have a hard time in the future.
Skim these
levelsio says Airbnb's whole rating scale is now 4.5 to 5.0
levelsio says Airbnb ratings now span 4.5 to 5.0 with a median of 5, and calls Yelp the only honest distribution.
Isenberg publishes a 13-step playbook for making physical products
Greg Isenberg's vibe manufacturing playbook: measure with calipers, have a model write the design as code, prototype in plastic, buy an engineer one hour, then send it to SendCutSend.
Frequently asked questions
What sales channel is Josh Pigford using for his unreleased product?
All of his sales in the first week came through Twitter direct messages, with no website or product launch yet.
How much revenue did Marc Lou make in September?
Marc Lou reported $217,071 in September across 17 products at a 90% margin, with HYROX bringing in $112K and his marketplace TrustMRR generating $53K.
Which AI model is now dominating code generation usage at T3 Code?
Opus 5.5 is now the first model to reach 50% of all prompts on T3 Code, with independent reports from multiple teams showing the same shift in the same week.
What did DHH discover when agents rewrote Campfire in different programming languages?
DHH had agents implement Campfire in Elixir, Go and Rust, and found that Ruby performed slowest compared to the other languages tested.
What is the new way to decide if a programming language will survive?
DHH and others argue that if AI agents cannot write a language well, the language's survival chances worsen, since factors like human enjoyment of writing in it matter less when agents generate most code.