Open field notebook with handwritten notes and annotations on a wooden desk
We're hiring — Feb 2026

What gets lost
when someone leaves?

The decisions nobody wrote down. The workarounds that became policy. The context that lived in one person's head for three years. Codex captures it.

12 min

avg capture time

94%

knowledge retained

3 days

faster onboarding

codex — new entry

→ Step 1 of 3: Role context

This helps Codex index the knowledge by function — so the next person in this seat finds it immediately.

This form is the product. No signup required.

tribal knowledgedecision historyinstitutional memoryunwritten rulesonboarding contextorg design rationaleprocess archaeologyrole-specific wisdomincident learningsculture documentationtribal knowledgedecision historyinstitutional memoryunwritten rulesonboarding contextorg design rationaleprocess archaeologyrole-specific wisdomincident learningsculture documentation

The knowledge-loss problem

The cost isn't the departure.
It's what leaves with them.

Scenario 01
Engineering Manager

Marcus's last day. Three years of decisions. Zero documentation.

Your senior backend engineer is leaving. He knows why the auth service was rebuilt in 2022, why you never use that third-party SDK, and why the deployment pipeline has that one step nobody touches. In two weeks, that knowledge is gone. The next engineer will spend six months rediscovering it through broken deploys and confused Slack threads.

Heard from

Engineering Manager, 60-person startup

Capture it in a 12-minute interview. Find it in 30 seconds.

Codex runs a structured exit capture with Marcus — not an exit interview, a knowledge interview. Every decision, every workaround, every 'don't touch that' gets indexed by role, system, and date. The next engineer finds it in their onboarding queue.

Dashboard showing captured knowledge entries organized by role and system with search functionality
codex — screenshot

47

Entries captured

12 min

Capture time

3 days

Onboarding saved

codex knowledge base
live
Scenario 02
People Ops Lead

The reorg erased the why. Now nobody remembers.

Six months after the restructure, your new team lead asks: 'Why does this process exist?' Nobody knows. The three people who designed it are in different departments. The Confluence page has seventeen edits and no rationale. The decision that made sense in Q3 2024 now looks like bureaucratic archaeology — and changing it means risking something nobody can name.

Heard from

People Ops Lead, 120-person SaaS company

Every process has a timestamp and an author. Every decision has a 'because.'

Codex attaches context to process, not just documentation. When you change something, you capture why the old version existed before you replace it. Future you — or future someone — will thank current you.

Timeline view showing process decision history with annotations and rationale for each change
codex — timeline

Why we split the on-call rotation

#process-history·8 months ago

The SDK we chose not to use (and why)

#decision-log·1 year ago

Deployment freeze — original rationale

#incident-learning·2 years ago
codex knowledge base
live
Scenario 03
Founder

Their tenth 'who do I ask about this?' Slack message. This week.

Your new hire is three weeks in. Brilliant. Motivated. Asking the same questions your last three new hires asked. The answers exist somewhere — in someone's head, in a Slack thread from 2022, in a doc that was accurate eighteen months ago. They're not failing to find answers. You're failing to put answers where they can find them.

Heard from

Founder, scaling past 50 employees

Onboarding that answers questions before they're asked.

Codex builds a role-specific knowledge queue for every new hire. On day one, they get the tribal knowledge for their seat — not a generic handbook. The questions they'd ask in week three are answered in week one. Your senior people stop being a help desk.

Onboarding dashboard showing new hire progress with pre-loaded role-specific knowledge entries and completion metrics
codex — metrics

23

Avg questions/week (before)

4

Avg questions/week (after)

6 hrs/wk

Senior time reclaimed

codex knowledge base
live

The people building it

We eat our own cooking.
Here's what only we know.

Every team member has a Codex entry. These are real entries from our team — the kind of thing that would have walked out the door if we didn't write it down.

Priya Nair, Head of Product at Codex, smiling in a bright office setting

Priya Nair

Head of Product

3 years at Codex

codex entryJan 14, 2026

Why we don't have a free tier

We tried it in 2024. Conversion was 0.3%. More importantly, free-tier users had no skin in the game and never completed a capture. The product only works when someone actually sits down and thinks. Friction is the feature.

#product-decision#pricing-history#lessons-learned

I want to work somewhere that treats knowledge like infrastructure.

James Okafor, Lead Engineer at Codex, focused expression in a technical workspace

James Okafor

Lead Engineer

2 years at Codex

codex entryFeb 3, 2026

The search index rebuild we never talk about

In November 2024 we rebuilt the entire search layer in a weekend. The old system couldn't handle entries longer than 800 tokens without degrading. We never announced it because we didn't want users to think we'd been shipping a broken search for six months. We had been.

#engineering#incident-history#infrastructure

The codebase has a memory. So should the team.

Sofia Reyes, Content and Ops at Codex, warm smile in a creative workspace

Sofia Reyes

Content & Ops

18 months at Codex

codex entryJan 28, 2026

The onboarding email that actually works

We A/B tested 11 variations. The winner had exactly 94 words, no bullet points, and ended with a question. The question was: 'What's the one thing your last company never wrote down?' Response rate: 34%. Everything else was under 6%. Nobody knows why. We stopped trying to understand it and just kept using it.

#content-strategy#onboarding#ab-testing

Good documentation is just good thinking, written down.

Every new hire at Codex adds an entry in their first week.

Open positions

Build this
with us.

We're a small team solving a problem that compounds with every hire, every departure, and every reorg. If you want to work somewhere that thinks about knowledge the way engineers think about infrastructure — we're looking for you.

EngineeringNew

Senior Full-Stack Engineer

Full-time
Remote (US/EU)

You'll own the capture pipeline — the core product experience where knowledge goes from someone's head to a searchable, structured entry. Think deeply about information architecture, not just data modeling.

TypeScriptNext.jsPostgreSQLLLM integration
Design

Product Designer

Full-time
Remote (US/EU)

The hardest design problem we have: making knowledge capture feel like a conversation, not a form. You'll work on the core flows that determine whether someone actually completes a capture or closes the tab.

FigmaInteraction designResearchSystems thinking
ContentNew

Content Strategist

Full-time
Remote

You'll build the knowledge taxonomy that makes Codex actually useful — the tagging system, the entry templates, the search vocabulary. You understand that how you label something determines whether anyone finds it.

Information architectureWritingTaxonomy designSEO
Engineering

ML / AI Engineer

Full-time
Remote (US)

We use LLMs to structure and tag knowledge entries, but we're not precious about which models or approaches. You'll run experiments, measure quality, and build the pipeline that turns raw captures into searchable structured knowledge.

PythonLLM fine-tuningRAG systemsEvaluation frameworks

The offer

We're a small team.
You'll feel it immediately.

No six-month roadmap handed to you. No feature factory. The work you do this month will be in front of customers next month. Your Codex entry on day one is the first thing the next hire reads.

Fully remote
Competitive equity
Learning budget ($3k/yr)
Async-first culture
4-day week (Fridays off)
Full benefits (US)

No cover letter required.
We'll ask you one Codex question instead.