It’s been a busy couple of months in the alpha. Most of what shipped was small and went straight into the changelog. Three changes are worth more than a line there, because they change what you see when you compare a resume to a job.
The loop is the same as ever: find a role, compare your resume against it, apply when the fit makes sense. What changed is how honest the comparison is, and how little you have to leave the job board to get it.
Comparisons run on newer models
Every resume comparison now runs on a newer generation of AI models than it did in July. Before switching, we replayed stored comparisons through the exact prompt the app uses and looked at what each model said.
The clearest difference showed up on a resume the app couldn’t actually read. Some PDFs are an image of a resume rather than text, and all the app can pull out of them is a title. The previous models sometimes scored that as a plausible fit and told the user to “review your resume first”. The new ones score it low, say skip, and say why: the file has a title and nothing else readable. That’s the answer you’d want from a friend who looked at it.
The comparison has always sorted gaps into two kinds: experience you don’t have, where it says “resume editing can’t close this gap”, and experience that’s probably there but not visible in the wording, where it suggests the bullet to add. The newer models are more consistent about which is which, and less likely to recommend a rewrite for a gap they’ve just called a real one. Knowing which kind of gap you’re looking at is most of the decision.
Two models are in use, matching the two comparison tiers:
- Standard comparisons, included with your plan, run on gpt-5.6-luna.
- Advanced comparisons, which use a credit, run on gpt-5.6-sol. It’s the more thorough of the two and takes a little longer.
On a job’s comparison view, the resume selector now shows which model each past comparison ran on, so a result from June and one from today are easy to tell apart. Older results keep their scores; we didn’t re-score anything.
Keep Apply To It open beside your search
The Chrome extension can now open in a side panel instead of the toolbar popup. The panel stays open while you move between postings, and you can resize it.
That changes the rhythm of a search session. Open the panel once, then click through listings on LinkedIn, Indeed, or ZipRecruiter. Each posting you land on can be captured and compared without reopening anything, so the question “is this one worth it?” gets answered while you’re still looking at the posting.
The popup is still there if you prefer it. The panel is for the long sessions.
A job view built around one question
The Job Details view in the dashboard and the in-extension job view were both rebuilt. The old layout showed you everything it knew about a posting. The new one is organised around whether the posting is worth your time, before you’ve even run a comparison.
The role summary, the posting’s stated requirements, and any concerns the app spotted in the posting itself (a stale or always-open listing, a missing salary, a visa or clearance requirement) sit at the top. The comparison, when you run one, reads as a guided path from the headline score down through the evidence, with the detail tucked into drawers you open when you want it.
In the extension the same view is leaner, because the extension is for deciding, and the dashboard is for digging in.
Smaller fixes you may have hit
- The extension no longer signs you out of the dashboard. The two were fighting over one session, and people were being logged out mid-search.
- Signing up with LinkedIn could leave you stuck on the setup screen with no way forward. Your name now carries over from LinkedIn, and skipping setup works.
- ZipRecruiter jobs track again from search results and company pages.
- When a job was saved without a captured description, the role summary now links you back to the posting to rescan it, instead of a dead end.
- On the Resumes page, the “used 0×” counter (which was never counted) and the bare version tag next to each resume are gone. The versions count and the version history stay.
What’s next
The next milestone is the beta. Between now and then the work is on the fundamentals: reliable capture across more job boards, comparisons that hold up across every kind of resume, and a clear path from “this looks like a fit” to the application itself.
If you’re in the alpha, the fastest way to shape that is to run a comparison on a posting you’re unsure about and tell us where the read felt wrong.