What each version added, and why. The badges below aren't screenshots: they're drawn live with the extension's own code and styles, so they show its real interface, which is in Chinese.
v1.8.0
2026-09-08New feature
You can hide companies you never want to see again.
Almost every similar tool abroad has a hide feature, and complaints about 104 on Dcard center on "seeing the same companies over and over". This is the only place the extension "decides for you", but the decision is yours: the tool only remembers and carries it out. It never judges on its own which company should be hidden.
① A "隱藏" (hide) button at the top right of the badge
It sits on this row instead of inside "說明" (details): people who hide usually clear out several companies at once, and one extra click is enough to make the feature disappear.
② Press it and that company's listings fold into one line
已隱藏:範例科技股份有限公司 — 你封鎖了這家公司
Not removed entirely. 104's search results use virtual scrolling, and pulling nodes out makes the scrollbar jump; and silently missing things are the hardest to debug. Leaving one line lets it say why, and lets you change your mind. 解除封鎖 (unblock) removes the rule; 顯示 (show) lets this one through just once.
This card matches a rule but is visible, so it was let through. Showing "隱藏" here would misreport the state: pressing it would do nothing and look broken.
Also
Hide listings I've applied to: needs you logged in to 104. Logged out, the data isn't available and the rule switches itself off rather than hiding the wrong things.
"N hidden on this page": shown in the toolbar menu. Without feedback, a filter just makes listings vanish and the tool looks broken.
Blocking uses the company code, not the name: renamed companies stay blocked, and different companies with the same name stay apart.
Deliberately not done: hiding promoted listings
Promoted placement is a product 104 sells. A switch that filters out paid exposure is close to ad blocking and would give them a solid legal reason to object. This extension fills in information 104 doesn't show; it doesn't interfere with how 104 sells ads.
v1.7.1
2026-09-08Fix
The toolbar menu still showed the previous version's grade table, which was wrong.
The menu said 0–19 正常 / 20–39 留意 / 40–64 可疑 / 65+ 高風險 (normal / watch / suspicious / high risk) and "scores are for reference only", but that whole system was removed in v1.7.0. Anyone opening the icon saw a feature that no longer existed. It now shows the three cues that actually appear on the badge.
The same version added a Buy Me a Coffee link at the bottom of the menu, not on the job cards.
v1.7.0
2026-09-05Major change
No more scores. The 0–100 score and four grades are gone.
All the data shows while collapsed; "說明" (details) only adds explanatory notes, no extra data.
They were removed not for lack of data, but because a test on 1,039 listings didn't support the deductions:
For listings up more than 60 days, 89.7% still had HR processing résumés within the week; the three over 180 days were 100%. Long-running listings are mostly always-open roles (building a talent pool, high turnover, tough requirements), not ghost jobs.
The batch flagged "no activity in 30 days" had a median posting age of just 1 day: new listings nobody had applied to yet, not ignored ones.
After 104 removed its HR activity field, the highest measured score was 30, while "suspicious" started at 40: the top two grades could never happen.
The green "opportunity" badge appeared on 33.3% of listings. One in three being an "opportunity" says nothing.
Flagging a real job as fake hurts job seekers and companies alike. Thresholds set by gut feeling can't get this right, so it now shows only the data and leaves the judgement to you.
Same version
Exact applicant counts are back: 104 shows a band like "6–10"; this shows "6".
The badge on job pages works again: 104's redesign moved the mount point and the job page had shown nothing.
Removed the store review prompt and its usage counter. Locally it now keeps only the API cache and listing observations, with no counts of any kind.
This version was also taken off the Chrome Web Store; see installing for why.