Uncountable Engineering  /  read this first

Before you interview, you should know this.

On November 13, 2026, Uncountable let go of 11 engineers, about a quarter of our engineering team.

We keep this page up on purpose. Anyone deciding whether to join us deserves to hear what happened directly from us, in full.

What happened

1. What happened?

The 11 people we let go were good engineers. They did solid work, and most had immense room to grow. Some had not been with us long enough to show everything they could do. The decision says nothing about their individual ability; it reflects a mistake in how we built the team.

The mistake was entirely ours. We planned engineering capacity the traditional way: estimate product scope, derive a budget, and convert that budget into headcount. That model holds when output scales linearly with engineer-hours. Over the course of 2026, that relationship broke.

By October 2026, agents authored 91% of our pull requests, and we merged 60,449 of them that month at a median open-to-merge time of 19 minutes. A year earlier, with essentially the same team size, we merged 1,885 PRs at a median of 16.5 hours (how we build →). These are not trivial or micro-scoped commits; they are substantive pull requests, comparable in scope to what our engineers authored manually when top contributors averaged 12 to 50 PRs a week.

Throughput grew, and our operational bottleneck moved. Engineers were no longer constrained by the hours they had to write code; they were constrained by available agent compute: tokens, concurrent sandboxes, and queue depth. Waiting for runs became the primary friction in an engineer's day.

Model usage and salaries come out of the same operating capital. Because we had allocated nearly all of it to fixed headcount, our engineers were starved of the compute required to run at full capacity. We recognized this structural leverage mismatch later than we should have, and we adjusted later than we should have. Rebalancing our capital toward agent compute meant letting good people go.

What's next

2. Could it happen again?

Yes. We would rather say that plainly.

We fully expect the role of the software engineer to change again, and we are looking for engineers eager to adapt as it does. Every new model generation alters how much leverage one engineer can direct, and no one can see more than a step or two ahead. That includes us.

Software engineering as a discipline will continue to shift, but what makes a great engineer has not changed, and will not change. It is still someone who stays close to the code, believes deeply in the craft, and cares relentlessly about good system design.

This is not a "vibe coding" job, and it will not become one.

Directing agents without understanding the underlying abstractions yields fragile software. Our durable competitive advantage is not prompt volume; it is the technical depth and rigor of our engineering team.

We intend to avoid another reduction through at least the next major jump in model capability. But we will not promise certainty past what we can clearly see.

Why join

3. Why join Uncountable now?

Because this is where the next version of the discipline is actively being forged.

A smaller team operating with maximum token efficiency and deep agent infrastructure gives each engineer here more leverage than almost anywhere else in software. You are not here to write features by hand. You are here to build the systems that build them:

Virtual file systems & sandboxing

Ephemeral, isolated runtimes that give agents reproducible execution environments at scale.

Model-agnostic coding harness

Our internal orchestration and development environment, evaluated continuously against real historical pull requests.

Agent knowledge layer

A specialized file and indexing layer that turns complex document management into data agents can read, update, and reason over.

Core entity engine

The extensible domain platform that serves as the foundation for both our scientific users and our automated agent fleet.

Fixes at this level compound immediately. A refinement to a review boundary, an agent skill, or an eval harness improves every run that follows. A good day does not merely close a single ticket; it permanently retires an entire class of bugs.

The stakes are real. Our software powers R&D for scientists inventing new physical materials, chemicals, and formulations. You are building the engine that accelerates their discovery.

The deal

What you are signing up for

You'll get You'll need to be comfortable with
Deep systems problems at the edge of how software gets built Operating without a static, multi-year roadmap
Substantially more compute and agent infrastructure per engineer than the industry standard Adapting your role as model capabilities evolve
Autonomous schedule flexibility: the agents run 24/7 so you do not have to Owning architecture well outside a narrow, defined lane
Colleagues who chose this high-leverage environment deliberately Automating your own work, repeatedly

If you want a predictable corporate environment and a stable year-long roadmap, this is not the right place, and that is a completely reasonable thing to want.

If the leverage and craft described on the left excite you more than the ambiguity on the right concerns you, we would like to meet you. We are hiring across our offices in San Francisco, New York, London, and Toronto for start dates in summer 2027. The company is growing rapidly, and we still need exceptional engineers to grow alongside it. We are bringing on far fewer people than originally planned, and the bar is high because each person shapes our trajectory.

Ask us about this page during your interviews. We hope you'll choose to join us, but we want you to do so with your eyes open to where software engineering is going, and ready for the ambitious work ahead here.

Jason HirshmanCo-Founder & CTO, Uncountable

Open roles

Eyes open? Apply.

San Francisco, New York, London, and Toronto.

Read how we build, or browse every open role.