查看他们最新的数据发布： https://t.co/3ZmuV5DxZm

Drops — MTS

Source: https://drops.mts.now/jobs/ Hiring snapshot · accelerator startups

Early Stage Hiring Patterns

We took a look at 910 graduates of the top accelerator programs from the past year and tried to figure out what they’re hiring for, how much they’re paying, and what stands out the most!

480Open rolesacross 114 companies

$170KMedian compp25 $120K · p75 $205K

36%Disclose comp171 of 480 roles

35%Mention equityin JD or comp summary

57%Engineering shareof all open roles

The market shape480 roles

Engineering alone is 57% of every open role. Outside of engineering, no single function dominates: ops, design, product, sales, and marketing each sit between 4% and 6%.

Compensation171 disclosed

Median midpoint is $170K across all disclosed roles.

Salary midpoint by functionn ≥ 10 disclosed

Salary midpoint by seniority

Top base salary by functionp75 of disclosed comp

Equity offered by function% of roles mentioning equity

Age cohort348 dated

Of the 480 roles, 348 are at companies whose YC batch we have on file. Winter 2026 startups (~4 months old) post fewer roles per company than 16-month-old Winter 2025 startups, but pay a higher median. Small samples per batch, so read it as direction.

Batch

Companies

Open roles

Roles / co.

Median comp

Top skill

Companies pre-Winter 2025 are in the dataset but absent here, since we only have batch records back to W25. The full snapshot above covers every role we found.

Stack276 eng roles

TypeScript leads at 26% of engineering roles, with React (15%), Python (14%), and Go (11%) clustered behind it. Rust shows up in about 1 in 17 listings — present, but a smaller slice than the frontier-model-era hype might suggest.

Share of engineering roles requiring this skill as a must-have.

How they work112 remote · 113 onsite

Of the roles that explicitly state a policy, remote and onsite are essentially tied (23% remote, 24% onsite, 2% hybrid). 51% of listings don’t specify at all.

Other dimensions

Visa sponsorship offered

16

3% of all roles · 0 explicit no

Median experience required

4 years

p25 3y · p75 5y · p90 7y

Comp-disclosing roles

36%

Ashby ATS: 61% · Greenhouse: 0%

Skills, all functionstop 30

The full demand picture, ranked. Junk tokens like ‘ai’, ‘rest’, and ‘analytics’ are filtered out. These are actual must-have skills extracted from JD text.

Methodology

Cohort

910 accelerator startups: every YC batch from Winter 2025 onward (672), all six Speedrun cohorts from a16z (231), and AI Grant (7). Neo and broader Crunchbase/TechCrunch-funded startups are excluded so the dataset reflects genuinely early-stage hiring, not late-stage names.

Hiring discovery

Probed /careers, /jobs, and related paths on each company, then detected the ATS by URL/iframe pattern. Of 858 reachable accelerator companies, 596 (69%) had some kind of public hiring presence.

Listings (two passes)

Structured: Ashby, Greenhouse, Lever, Rippling public APIs. Homegrown careers HTML scraped + parsed by Haiku agents for everything else. End result:480roles across114companies.222additional companies had careers pages with no actual listings on them.

Extraction

Eight parallel Claude Haiku 4.5 agents read every job description and extracted comp, equity, skills, seniority, remote policy, and years of experience. Comp ranges from Ashby and Lever are pulled from their native structured fields when available.

Caveats

• **The companies missing from this page.**Of the 858 reachable accelerator companies, 262 (31%) have no /careers, no /jobs, and no public ATS, so they hire through their networks or not at all. Of the 596 who do have hiring infrastructure, only114list specific open roles; the rest just have a “we’re hiring” placeholder.
• **Comp disclosure correlates with ATS choice.**Companies on Ashby disclose comp 61% of the time. Companies on Greenhouse: 0%. That’s not because Greenhouse companies pay worse, it’s because Ashby ships a structured comp field and Greenhouse doesn’t.
• **Skill extraction is approximate.**The model decides what counts as a “skill”. We filter the worst offenders but the long tail of niche skills will undercount.
• **Snapshot, not trend.**Data was pulled on2026-05-22. It will go stale.