The resume that updates itself

Every resume is already wrong.

Not wildly wrong — the facts are probably accurate. But a resume is a snapshot of the moment you wrote it. The day you ship the feature that solves the thing you were hired to solve, nothing updates. The day you learn Rust by actually shipping production Rust, nothing updates. The day you lead your first incident response and your team calls it the most organized RCA they’ve seen — nothing updates.

Your resume is static. Your work is not.

The problem isn’t laziness

People update their resumes when they need them. Usually when job-hunting, occasionally after a big promotion. Not continuously.

This isn’t laziness. It’s that the maintenance cost of continuous updating is too high to justify. Adding a bullet point to a Google Doc every week for two years before you need it is a discipline almost nobody sustains.

So everyone does the same thing: when the job search starts, they dedicate a weekend to reconstructing three years of context from memory, Slack search, old Jira tickets, and the occasional “what did I actually do that week?” panic.

The reconstruction is both painful and lossy. Things that would have been great resume bullets are gone — not because they weren’t real, but because you didn’t write them down in the moment.

GitHub is a log of your work

If you’re a software engineer who ships code, your GitHub commit history is a structured record of most of your professional output. Not all of it — whiteboard decisions, architecture reviews, production incidents, mentoring — but a lot of it.

Each commit has a date, a message describing what you did, and the diff of what changed. With a language model, a git log is a CV waiting to happen. Not verbatim — “fix typo in README” doesn’t go on a resume — but summarized, filtered, and transformed into achievement-language.

This is what EverCV does. Connect your GitHub account, and overnight the system reads your recent commits, filters out the noise, and updates your public CV page with the meaningful work.

What actually happens

The nightly cron fetches new commits, filters merge/bump/noise commits, groups by repository and week, and passes the grouped commits to Claude: “Summarize these commits into achievement-language for a senior engineer’s CV. Be specific. Drop the noise. Write like the person, not like a job posting.”

The output:

Shifted the recommendation engine from a single global model to per-user fine-tuned variants (3 repos, 47 commits, Jun 2–8) — Rewrote the training pipeline to shard by user cohort, reducing cold-start latency from 2.3s to 180ms. Migrated the feature store to DynamoDB on-demand to handle the 3× increase in write volume.

That’s not from a commit message. It’s a synthesis of 47 commit messages across three repos in one week. The numbers come from the commits; the framing comes from Claude.

The part that surprised me

I expected the value to be convenience. The actual value, when I started running it on myself, was something different: I caught things I would have under-described.

I shipped a performance optimization last February. I would have written it as “optimized query performance” and moved on. The commit messages had the actual numbers — 2.3s to 180ms, 3× write volume. The CV entry has those numbers. That’s a meaningfully different sentence.

Engineers are bad at self-promotion. Not because they’re modest — because the technical details that make the work impressive are in the commits, and pulling them out of 47 commits is work. The system does that work automatically.

Who this is for

EverCV is for engineers who push code to GitHub regularly, are in or about to enter a job search, and want a living record of their work without maintaining it manually.

$15/month (Prosumer) for continuous updates. $39/month (Pro) for richer summaries including PR discussions (coming soon). Free tier for a one-time 90-day snapshot.

evercv.io — launching soon, join the waitlist.