Where did the frontend jobs go? A year of Jobinja listings, measured
Frontend postings on Jobinja fell 30% in one year — three times faster than their category and seven times faster than the market — while AI postings grew 41%. Here's the data, the method, and the caveats.
Everyone in Iranian tech Twitter has a feeling about the job market. Feelings are cheap. Numbers are better.
So I measured it. Jobinja shows an exact count of active listings on every search and category page — “۳۵۱ فرصت شغلی فعال یافت شد” — and the Wayback Machine happens to have archived many of those pages last summer. Compare the archived count with today’s count on the identical URL, and you get a clean year-over-year measurement: same page, same query, same counting logic.
The headline: frontend developer postings fell 30% in one year. The market as a whole fell 4.5%.
The numbers
Each row compares a Wayback Machine snapshot from 2025 with the same URL fetched live on August 16, 2026.
| What was measured | 2025 snapshot | 2025 | 2026 | Change |
|---|---|---|---|---|
| Frontend search (react / front / فرانت, web dev category) | Aug 10, 2025 | 351 | 244 | −30.5% |
| Web, programming & software (category) | Aug 10, 2025 | 2,259 | 2,031 | −10.1% |
| AI search (هوش مصنوعی) | Jun 28, 2025 | 228 | 321 | +40.8% |
| Python (keyword search) | Aug 6, 2025 | 257 | 284 | +10.5% |
| DevOps: IT/DevOps/Server (category) | Jun 28, 2025 | 815 | 625 | −23.3% |
| Design (category) | Jul 25, 2025 | 612 | 438 | −28.4% |
| Graphic designer (role page) | Jul 7, 2025 | 441 | 248 | −43.8% |
| Sales specialist (کارشناس فروش, role page) | Jul 6, 2025 | 2,711 | 2,852 | +5.2% |
| All active listings on Jobinja | Aug 6, 2025 | 15,043 | 14,372 | −4.5% |
The frontend row is the cleanest measurement in the table: its baseline is August 10, 2025 — an almost exact one-year window — and it comes from a specific search (react front فرانت) scoped to the web development category.
Three things the data says
1. Frontend fell three times faster than its own category. The web & software category lost 10% of its listings; frontend lost 30%. The overall market lost only 4.5%. So this isn’t “Jobinja shrank” or “tech shrank” — frontend specifically cooled.
2. Demand is rotating, not just shrinking. Inside the same shrinking category, AI-keyword postings grew 41% (and that’s against a June baseline; against a May baseline of 267 they still grew 20%). Python grew 10%. That’s a ~70-point spread between AI and frontend within one category, in one year. The pie got slightly smaller; the slices moved a lot.
3. Frontend is not uniquely cursed. Graphic design postings fell 44% and the design category fell 28% — as hard or harder than frontend. Sales grew 5%. The pattern reads less like “frontend is dying” and more like a broad pullback in UI-adjacent and creative hiring, with frontend as the soft spot within software engineering: without it, the rest of the web & software category would be roughly flat.
What I couldn’t measure
Honesty section. Some numbers I wanted don’t exist:
- Backend and full-stack have no 2025 baseline. The only archived backend search snapshot is a 1.7 KB anti-bot challenge page with no content, and no full-stack page was archived at all. Today backend and frontend keyword searches sit at near-parity (226 vs 223 listings) — but I can’t tell you which direction backend moved, only that its two rough proxies point opposite ways (Python +10%, Django −53% on a stale 15-month baseline).
- A few baselines are late-June/July rather than August, so those rows are ~13-month comparisons, not exactly twelve.
- One tempting number was excluded. The sales & marketing category shows −41.8%, but its only usable snapshot is May 2025 — fifteen months back, not year-over-year — and it flatly contradicts the sales-specialist page (+5.2%). Seasonal confound, most likely. Into the bin.
Method and caveats
The method is reproducible in an afternoon: find the page’s Wayback Machine snapshot, read the listing count from the archived HTML, fetch the same URL live, read it again. No scraping of individual postings, no estimation — Jobinja does the counting on both ends.
The caveats are real, though:
- Keyword search is a proxy. The frontend query catches some full-stack postings and misses ones that name only a framework. The same bias applies to both years, which is why the change is meaningful even if the absolute count is fuzzy.
- Counts fluctuate week to week. The software category read 2,259 in August 2025 but 2,901 in an October 2025 snapshot. Single-snapshot comparisons carry noise; the frontend signal (−30%) is well outside it, the market signal (−4.5%) is not.
- Postings measure advertised demand on one platform, not hires, not salaries, and not the whole Iranian market.
If you want to check my work, the key snapshot is the frontend search as archived on August 10, 2025 — run the same search on Jobinja today and count.
So what do you tell a junior frontend developer?
Not “learn a different job.” The honest read is narrower: the entry ticket that worked in 2021 — React alone — is the part of the market that’s evaporating, and the listings that grew are the ones where frontend skills sit next to something else: AI integration, Python, product breadth. The market didn’t stop hiring people who build interfaces. It stopped hiring people who only build interfaces.
Build it, don't just read about it
The 12-week AI Engineering mentorship takes you from an empty file to a complete AI coding agent — every subsystem, real code, no frameworks.
Six tools and a lot of caution: reading Claude Code's tool layer
Your agent's six tools are 600 lines. Claude Code's are thousands — and the whole difference is edge cases somebody watched break. Fail-closed defaults, an A/B test they ran and reverted, curly quotes, why Bash(python:*) is an off switch, and the tool list that's sorted for the cache.
Sandboxing: where your agent's code is allowed to run
Letting the model write code is the highest-leverage tool you can give an agent — twelve tool calls collapse into one. It also turns text your agent read into a program running in your process. Thirty lines of node:vm, and the one idea that survives every upgrade to real isolation.
Reading the real Claude Code: six files that teach you the REPL
Claude Code's REPL is 5,005 lines. You should not read it. Here are six small files — streaming, slash commands, abort trees, cancellation, and the cost meter — that carry the actual lessons, with the code inline and every line linked.