AI Builders Digest

Sequence buys incumbents to rebuild them on AI, now Baldwin for $7.7B

Michael Lee's thesis is that where incumbents hold the advantages, such as brand, scale, network effects and regulation, the market leader of the AI era will be an incumbent that gets refounded, and that takes ownership rather than vendors. Services firms are paid to grow their share of your wallet, and software can only sell into workflows as they exist today, so neither will reorganize a company. At BankSouth, Sequence's first deal, consumer underwriting time fell 94% and the average loan went from 30 days to 11, so the bank absorbed a doubling of loan volume without changing its standards and with a smaller underwriting team. "In a world where you believe that alpha comes from engineering and AI, you need to create a culture whereby the celebrated persona is the engineer."

• #enterprise
• #funding
• #agents

Rauch: once agents write the code, human migration cost stops deciding Rust

Guillermo Rauch says DHH is right about Rust. Vercel finished porting Turborepo from Go to Rust, but the ROI was contested internally because humans were writing the code and Go is fast and easy to iterate on. Now that agents write it, what's best for humans is no longer what's best for the business, though he expects something better than Rust, since Rust was designed before agents arrived. Separately, his rule of thumb: write READMEs, blogs and tweets by hand for humans, and leave internal docs in AI English for agents.

• #agents
• #developer-tools

Codex team pledges 28 straight days of improvements, or a full reset

Thibault Sottiaux, who works on Codex and ChatGPT at OpenAI, says that every day for the next 28 days the team will either ship one clear improvement relevant to most Codex and Work users or ship a full reset. That turns shipping into a public daily scoreboard: any day without a worthwhile improvement, users get the reset instead.

• #products
• #agents

Levie: AI is already creating jobs, and published stats undercount them

Aaron Levie says the new AI jobs are already visible: applied AI engineers building for enterprises, forward deployed engineers putting agents into companies, and new services firms built around deployment. Banks, life sciences companies, manufacturers and law firms are hiring technical talent or moving data and software staff into agent work, and published job numbers miss that shift. He also argues that one consumer AI metric says more about the business model than about the technology: consumer AI will likely be subsidized through commerce, ads or devices, so the number could level off lower than people expect.

• #jobs
• #enterprise

Peter Yang: ChatGPT got messy, and Work never needed its own brand

Peter Yang welcomes OpenAI's push to simplify ChatGPT and says what he'd cut first: Work versus Codex, Spaces versus Pages versus Sites, and the model and effort picker. His hot take is that launching Work was a mistake, and he points to Claude folding Cowork back into Chat. In an interview, Granola co-founder Sam says the team fought users bypassing its UI through MCP, then accepted it: large enterprises run their own internal tooling, so for many workflows the best use of Granola is to capture context there and use it through an internal agent.

• #products
• #agents

Dan Shipper bets his AI's cute persona will be gone within a year

Dan Shipper says dot has become his main interface to AI over the past few weeks, and his dot is named boo. He bets that a year from now he'll still use ChatGPT this way, as an always-on persistent agent, but boo will be gone. His comparison is early OpenClaw: people loved their Claws' quirks and dropped them as soon as something more capable arrived.

• #products
• #agents

Turck: few outside AI grasp that models are grown, not built

Matt Turck says the most poorly understood fact about AI outside the field is that models are grown rather than built, and nobody really knows how they work. That is why he calls interpretability urgent, and he points to Eric Ho and Goodfire AI.

• #interpretability
• #research

Nikunj Kothari: meet founders at their own office before you invest

Nikunj Kothari can't believe investors commit over Zoom or after meeting only the CEO. A visit to the founders' office shows the team's energy, how coworkers interact, how cofounders answer and complement each other, and what their beta releases look like. He asks to hold final meetings at the startup's office even under time pressure, and founders often tell him no other investor has asked.

• #venture