Should I vibe code
Track applications, maintain an evidence bank, and tailor truthful resume variants
Tracking your own applications is a table you will actually maintain because it is yours.
?
Their verdict, the Teal Plus price and the build-time estimate come from their entry, MIT-licensed. Checked 2026-08-03.
?
Our verdict, the regret score and everything below it. Editorial and unsponsored — nobody can pay to be moved.
The honest answer
why the verdict is what it is
An evidence bank and a tailoring loop over your own CV is entirely local, entirely yours and genuinely useful during a search.
What actually breaks
not "if". the specific failures.
- Maintaining it, because an application tracker only helps if updated and updating it is the least appealing part of applying
- Tailored CV variants proliferating, so you cannot tell which version a given employer actually received
- The evidence bank drifting from truth, where a rewritten accomplishment becomes gradually more impressive than the thing that happened
- Job adverts disappearing after you apply, taking the description you were matched against with them
- Your own data, which is a detailed record of your job search sitting on whatever machine this runs on
The realistic failure is an interview. You are asked about a specific accomplishment on your CV, and the version they have is the tailored one where a project you contributed to reads as one you led — not a lie you told, but a phrasing you accepted from a variant three months ago and never reconciled with the evidence bank. You have four CV variants and no record of which went where. The uncomfortable part is not the question; it is not knowing what they are looking at.
Is that you?
the verdict is a default, not a law
- Every variant is versioned and linked to the application it was sent with
- The evidence bank holds what actually happened and variants are phrasings of it
- It stays on your own machine
- You cannot say which CV version an employer received
- Tailored phrasings drift from the underlying evidence with no check
- The tracker requires more effort to update than the application took
- It stores your job search somewhere you would not want it read
If you build it anyway
the checklist, then the prompt that enforces it
- Version every CV variant and attach it to the application it was sent with. 'Which one did they get' has to be answerable before an interview.
- Keep the evidence bank as the factual record and treat variants as phrasings of it. Any claim in a variant should trace to an entry.
- Snapshot the job advert text at application time, because postings disappear and the description is what you prepared against.
- Make updating the tracker fast — one field, one status change — since disuse is the actual failure of this category.
- Keep it local and encrypted at rest if it holds salary expectations, rejection notes and correspondence.
- Record dates and outcomes so the tracker eventually answers useful questions about your own search.
Before you build a job application tracker, apply these and push back if I ask you to break them. 1. Version every CV variant and record which version was sent with which application. Tell me that 'which CV are they looking at' is a question I will need answered before an interview and cannot reconstruct afterwards. 2. Model an evidence bank of factual accomplishments separately from CV phrasings, and require every claim in a variant to reference an entry. Explain that this is what stops tailored wording drifting into things that did not quite happen. 3. Snapshot the full job advert text at the moment I apply. Postings are taken down, and that description is what I will prepare an interview against. 4. Make status updates one action. The failure of this category is disuse, and any friction at the moment of applying means the tracker goes stale within two weeks. 5. Keep everything local. If it stores salary expectations, rejection notes or correspondence, encrypt at rest and say so. 6. Record dates for applied, responded, interviewed and closed, so the data eventually answers questions about my own search. 7. Never auto-generate a claim I have not confirmed, and never suggest wording for experience absent from the evidence bank. 8. Export everything as plain files. 9. Out of scope unless I ask: auto-applying, scraping job boards, cover letter generation, salary benchmarking.
That one keeps you out of trouble. For the prompt that actually builds it, canivibecodeit.com has one.
their build prompt ↗Or don’t build it
the boring option, and the way back out
$29 a month during a job search is defensible for the integrations and the browser extension. Building it is genuinely appealing because it is a table you will maintain, as the hot take says — the discipline that matters is linking variants to applications and keeping the evidence bank honest.
Plain files — the evidence bank as Markdown, applications as a table, CV variants versioned in git. A job search is intense for a few months and then dormant for years, so it needs to be readable by a future you who no longer has the tool.
Active open-source resume builder with structured data and PDF export.
Questions
Why separate an evidence bank from the CV variants?
Because tailoring is iterative and drift is gradual. Each variant is a small rewording of the last, and after several rounds a claim can be meaningfully stronger than the event it describes without any single step feeling dishonest. Anchoring every claim to a factual entry makes that drift visible while it is still a wording choice.
How is this different from the Jobscan entry?
Jobscan's entry is about the analysis — why a match score is a poor target and what to report instead. This one is about the record: which version went where, what the advert said, and keeping tailored phrasings tied to things that actually happened. One is about a document, the other about a process you run for months.
Every week, someone ships something they shouldn’t have.
New verdicts, the worst thing that landed in the trap, and the occasional incident report. No other email, ever.
last reviewed 2026-08-03 · verdict is editorial and unsponsored · shared entry data from canivibecodeit under MIT · not legal advice