I was looking for a new role. Like many engineers, I have specific requirements:
- Remote-friendly: I want to work from home, or at least have hybrid options
- Senior-level: 5+ years experience, or titles like Senior/Staff/Principal
- Salary disclosed: I don't want to waste time on roles that pay below market
The challenge? Job boards are noisy. Hacker News "Who's Hiring" threads contain hundreds of postings per month, but most don't meet all three criteria. Manually scanning each one takes hours.
I used Claude Code to pull 1,000 job postings from the last 5 months of "Who's Hiring" threads via the API. Now I had a spreadsheet, but how to filter it?
The regex trap
The obvious first approach is keyword matching:
# Naive approach
is_remote = "remote" in description.lower()
is_senior = "senior" in title.lower() or "5+ years" in description
has_salary = "$" in description
This fails badly:
- "No remote work available" contains "remote" but means the opposite
- "Senior year intern" contains "senior" but isn't a senior role
- "$0 in funding" contains "$" but isn't a salary
- "Competitive salary" or "DOE" technically mentions compensation but doesn't disclose it
A regex-based approach achieves 68% precision, leaving hundreds of false positives to manually review.
So I tried something different: I ran the 1,000 postings through everyrow.io/screen with these criteria:
Qualifies if ALL THREE are met:
- Remote-friendly: Explicitly allows remote work, hybrid, WFH, or distributed work
- Senior-level: Title includes Senior/Staff/Lead/Principal/Director/Architect, OR requires 5+ years experience
- Salary disclosed: Specific numbers provided (e.g., "$150K", "$120-180k")
Mark False if: On-site only, junior role, "competitive" salary, or any criterion unclear
What survived
| Metric | Value |
|---|---|
| Total screened | 1,000 |
| Qualified | 148 |
| Pass rate | 14.8% |
| Processing time | ~20 minutes |
| Cost | $2.90 |
So roughly 1 in 7 HN job postings meet all three criteria. Not bad!
Pass vs. fail
Spot-checking some of the results, it seems to be working as expected
Adobe | Senior Software Engineer | $173,500 - $331,050 | Remote-friendly
StartupCo | Staff Engineer | San Francisco (Remote OK) | "Competitive compensation"
BigCorp | Senior Developer | $180,000 | New York (Must be in office)
TechCo | Software Engineer | $90,000 - $110,000 | Remote
Is it worth it?
| Approach | Cost | Time | Precision |
|---|---|---|---|
| Regex filtering | Free | 20 hrs to build | 68% |
| Manual review | $125 @ $25/hr | 5 hrs | ~95% |
| DIY LLM API | ~$4 | 20+ hrs dev time | ~80% |
| everyrow.io/screen | $2.90 | 6 min | >90% |
The DIY approach is tempting, but a naive implementation actually costs more—everyrow.io uses batching and prompt optimizations that most one-off scripts skip. Add the engineering time for consistent schema, error handling, and retries, and it's not close.
The same approach works for any filtering task where keywords fail you: lead qualification, content moderation, investment screening, supplier vetting. Try it yourself, the free tier covers 50 rows.
Data: 1,000 job postings from Hacker News "Who's Hiring" threads (August 2024 - January 2025). Session URL: everyrow.io/sessions/aecfddce-da1a-43b3-ad15-a05af0a9ae72