Scuttlebutt

Scuttlebutt

SaaS will be one of the largest winners of the AI game — TypeSafe AI's founder

13 takeaways: why coding agents write the same software faster, why OpenAI has tried to automate customer service since 2020, and why reliability is the whole product

Scuttlebutt's avatar
Scuttlebutt
Sep 29, 2026
∙ Paid

The a16z Show · Sep 28, 2026 · 42m · 29m saved

With Diogo Almeida, Co-Founder and CEO of TypeSafe AI

SaaS will be one of the largest winners of the AI game — TypeSafe AI's founder

OpenAI has been trying to automate customer service since 2020, Diogo Almeida said, and AI that can pass graduate-level science tests still cannot run a drive-through.

Coding agents such as Claude Code and Codex write software faster, but the code they produce is the same kind people wrote a decade ago. Almeida's company, TypeSafe AI, is betting on the opposite: a model that sits inside a program and makes decisions the program can act on.

"What I want instead is smart software."

Diogo Almeida, co-founder and CEO of TypeSafe AI, on The a16z Show, worked at Google Brain and at OpenAI, where he helped release the model trained with human feedback that came before ChatGPT. TypeSafe's first model, Jev, turns plain-language instructions into choices a program can use.

The full interview is covered here so you can skip it. 42 minutes of audio, 13 minutes of reading.

Here are the 13 takeaways that matter.


This Scuttlebutt Summary is available in full to premium subscribers. If you're not one already, please consider subscribing to keep reading.


Key Takeaways

  • Coding agents make the same software faster, and the software itself is not getting better

  • Jev is a classifier a programmer puts inside the code, not a tool that writes code

    • It takes a description in plain language and a set of states, and returns a choice with a confidence level

  • Intelligence per dollar is the design goal, even if speed matters more in the short run

  • The industry optimized models to impress human judges instead of to automate work

  • Reliability is the product: similar intelligence every time, not identical output

  • Coding agents are good at syntax and "incredibly bad at architecture"

  • The sell-off in software-as-a-service stocks got it wrong, he said, and SaaS companies are best placed to automate their customers' workflows

  • Most calls to AI will eventually happen deep inside software, not in front of a person

1. Where Is the Automation

Asked for an elevator pitch, Almeida said his favorite is a blunt question: where is all the automation? AI is extremely smart, he said, yet almost useless outside chatbots and coding agents, which he uses and likes.

"But TypeSafe is making AI for software." — Diogo Almeida

He said TypeSafe wants AI to be powerful for building real software, not only with a person in the loop, and that Jev is its first model. One of the hosts said Jev had "caught fire" among developers, who were calling to say how good it was.

2. Not Just Faster Code

User's avatar

Continue reading this post for free, courtesy of Scuttlebutt.

Or purchase a paid subscription.
© 2026 The Golden Age LLC · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for great culture