> Markdown version of [/magazine/783-infer-once-run-everywhere-zero-token-architecture](https://www.wearedevelopers.com/magazine/783-infer-once-run-everywhere-zero-token-architecture). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Infer Once, Run Everywhere: Zero-Token Architecture

**By:** [Daniel Cranney](https://www.wearedevelopers.com/@daniel-cranney)

**Published:** October 13, 2026

> “You are all having way too much fun burning way too many tokens”.

Kelsey Hightower kicked off his talk at the WeAreDevelopers World Congress North America by calling out the developers in attendance for something so many of us are guilty of, even if we don’t realise it.

What if, before we needlessly pass a repeatable task - end to end - to an LLM, we take an approach that might feel a little less cutting-edge in the _agentic-everything_ era, but an approach that might save an enormous amount of wasted tokens, energy, time and, of course, cost.

### Zero Token Architecture (ZTA)

The term ZTA - coined by Kelsey “out of spite” after seeing posts from users on LinkedIn bragging about agent counts - has a simple idea behind it: **infer once, run everywhere**.

Essentially, use AI to work something out, then run the result as ordinary, predictable code.

An agent is a complex mix of a model, a prompt, tools, an MCP server and skills, and running inference on each step goes against the basics of automation.

There is a security cost too, software that can run arbitrary code was once called malware, and the industry spent years working towards zero trust, but as Hightower reminded our attendees in San Jose, it’s easy to forget all that because the new tool is exciting.

## Things to Try

1. **Question the loop.** Does this task need inference every time it runs, or only once to create it?
2. **Build with AI, then export as a script.** Hightower’s suggested final prompt: turn all of that into a shell script and remove the dependencies you don’t need.
3. **Let a scheduler do the repeating.** The tools already on your machine can run it for free.
4. **Prove it with numbers.** Time your sessions and work out what each loop costs, then compare after the export.
5. **Keep your judgement.** Don’t paste chatbot output into your terminal, or hand personal decisions over to it, stay in the loop.

## Watch Kelsey Hightower from San Jose

You can watch Kelsey’s talk ZTA: Zero Token Architecture, recorded live at the WeAreDevelopers World Congress North America in San Jose, in-full below, or [check it out later here](https://www.wearedevelopers.com/videos/100533-zta-zero-token-architecture).

https://www.youtube.com/embed/https://youtube.com/live/WtPoI8jrBLs


## Related Articles

- [Never delegate the understanding](https://www.wearedevelopers.com/magazine/749-never-delegate-the-understanding)
- [Dev Digest 132 - Binging WADFlix?](https://www.wearedevelopers.com/magazine/473-dev-digest-132-binging-wadflix)
- [The Web We Broke (And Why AI Agents Are Paying the Price) - AgentCon Berlin](https://www.wearedevelopers.com/magazine/735-the-web-we-broke-and-why-ai-agents-are-paying-the-price-agentcon-berlin)
- [Dev Digest 137 - AI'm not sure about this](https://www.wearedevelopers.com/magazine/485-dev-digest-137-ai-m-not-sure-about-this)

## Related Videos

- [ZTA: Zero Token Architecture](https://www.wearedevelopers.com/videos/100533-zta-zero-token-architecture)
- [Ship Smarter Agents, Not Bigger Prompts](https://www.wearedevelopers.com/videos/100010-ship-smarter-agents-not-bigger-prompts)
- [Agentic employees in world's most downloaded FinTech app](https://www.wearedevelopers.com/videos/100123-agentic-employees-in-world-s-most-downloaded-fintech-app)
- [From minutes to seconds: Lessons learned optimizing multi-turn agentic workflows](https://www.wearedevelopers.com/videos/2078-from-minutes-to-seconds-lessons-learned-optimizing-multi-turn-agentic-workflows)

## Related Jobs

- [Founding Developer Advocate](https://www.wearedevelopers.com/jobs/48425-founding-developer-advocate) at **Plumber**
- [Software Engineer](https://www.wearedevelopers.com/jobs/ext/2596233-software-engineer) at **Twilio**
- [Software Engineer - Video](https://www.wearedevelopers.com/jobs/ext/2600051-software-engineer-video) at **Twilio**
- [Software Engineer, Fullstack](https://www.wearedevelopers.com/jobs/48415-software-engineer-fullstack) at **Sciforium**
- [LLM Training Engineer](https://www.wearedevelopers.com/jobs/48420-llm-training-engineer) at **Sciforium**
- [Software Engineer (L3)](https://www.wearedevelopers.com/jobs/ext/3535339-software-engineer-l3) at **Twilio**