SignalFire's June 24 Report: The Best New Grads Are Founders, Not Applicants

SignalFire dropped its 2026 State of Tech Talent Report the week of June 23, and TechCrunch and Channel Dive both led with the same scary number: entry-level hiring at major tech companies is down 65% since 2019. The scarier read, if you are trying to hire new grads in 2026, is what the top quartile of those grads are doing instead. They are not unemployed. They are not waiting. They are founders.

If your new-grad pipeline still starts at a campus career fair, a Handshake job post, or an ATS funnel, you are fishing in a pond the best fish already left.

The 65% drop is a redirect, not a vacuum

SignalFire's headline finding (analysis across millions of employees and more than 80 million companies) is that the entry-level seat at the 12 Tech Majors has collapsed. Alphabet, Meta, Apple, Amazon, Microsoft, Netflix, Nvidia, Tesla, Uber, Airbnb, Block, and Stripe have cut entry-level hiring 65% since 2019. Early-stage startups have cut it 75%.

But total engineering hiring across those Tech Majors is only down 11% versus an overall headcount drop of 25%. Engineers went from 46% of new hires in 2019 to 55% in 2025. The work didn't disappear. The first-year work did.

The 12 to 18 months a junior used to spend writing boilerplate, running unit tests, and debugging under a staff engineer's eye? That's now Copilot's job, Cursor's job, and a Claude Code subagent's job. So the Tech Majors stopped hiring for it. And SignalFire found that the seats labeled "junior" that are open are increasingly filled by laid-off senior ICs with 10+ years of experience. A new grad in 2026 is not competing with their classmates. They are competing with a former staff engineer from Meta who'll take a pay cut for a year of stability.

2x

This is the part that matters for sourcing. The new-grad founder boom is partly opportunistic (cheap compute, AI-native fluency, capital flowing toward solo and pair founders) and partly defensive. If your only options at 22 are "compete with an ex-Stripe staff engineer for a Level I seat" or "start something," a lot of Stanford and MIT graduates are picking option two. Asher Bantock, SignalFire's head of research, framed the layoffs rationale as "consistently AI" but pointed out the hiring data contradicts the story. The seats aren't vanishing. They are reallocating up the seniority curve.

Who the "top CS grads" actually are

One easy mistake when reading the SignalFire numbers: the "top CS grads" group is not Ivy League. It is the top 20 undergraduate engineering programs per U.S. News. That puts CMU, Georgia Tech, UIUC, Michigan, UT Austin, Purdue, and Berkeley in the cohort alongside MIT and Stanford. The implication for sourcing new grad engineers is that the founder-redirect is happening at scale across a dozen feeder schools, not just two coastal brands.

And the archetypes back this up. Carina Hong dropped out of Stanford to build Axiom Math, an AI mathematician that has already solved two Erdős problems and raised a $64M seed. Brendan Foody dropped out of Georgetown to start Mercor. Michael Truell finished MIT and went straight into Cursor. Scott Wu finished Harvard and went straight into Cognition. Two Stanford students, Roman Scott and Itbaan Nafi, raised $2M from Mayfield and Collide Capital for Breakthrough Ventures, a fund whose entire thesis is bankrolling other college and new-grad founders nationwide.

The capital infrastructure is now pointed at the alternative path. That is what "redirect, not vacuum" looks like in practice.

Why your sourcing tools fail on this cohort

Here is the operational problem. New-grad founders do not describe themselves the way your boolean searches assume. Run a class-year keyword search for "Class of 2024" or "Class of 2025" founders in the U.S. against any professional-network index and the aggregate comes back near zero. We checked. These people do not put a graduation year in their LinkedIn headline. They put "Building Axiom." "Founder, stealth." "We're hiring." "ex-Stanford."

A recruiter searching for "new grad" or "entry level" or "Class of 2025 software engineer" is querying a column that the actual target cohort doesn't fill in. Meanwhile the signals that do identify them sit across three or four different surfaces:

• A GitHub profile with heavy 2024 to 2026 commit activity on a public repo with real stars and real users.
• A LinkedIn that says "Founder" with a one-month tenure and a school in the experience section dated 2021 to 2025.
• A YC W26 batch listing, a TreeHacks 2025 demo page, a Free Ventures cohort directory, a StartX or Cardinal Ventures profile, an MIT Sandbox grantee list.
• A personal site with a "now" page and a Twitter / X account where the bio reads "building [thing] @ [school]."

No single ATS field captures this. No single search filter captures it. This is exactly the friction we built Refolk to remove: you describe the person in plain English ("CS grads from a top-20 engineering school who founded something in the last 18 months and are still actively shipping on GitHub") and you get a ranked shortlist across GitHub, LinkedIn, and the open web, instead of three separate boolean queries that each miss two-thirds of the cohort.

The artifact is the new GPA

Tawni Cranz, quoted in the Channel Dive coverage on June 26, put it bluntly: "I think new grads need to be building artifacts and prototypes versus just graduating from the schools and having the skill sets." She was talking to candidates. The mirror statement, for recruiters and founders, is that the artifact is now the highest-signal column on the candidate record.

GPA was a proxy for "this person can grind through abstract work." A shipped product with users is the actual thing. A 200-star repo with a clean commit graph since sophomore year tells you more about whether someone can ship into your codebase than any internship at a Tech Major can, because internships in 2026 are also Copilot-assisted boilerplate work.

The "junior seat" got automated. The "junior founder" seat got funded. Your sourcing layer has to follow the money.

This is why a smart entry level tech hiring 2026 strategy starts from the artifact side. Pull the candidate from their GitHub graph, their hackathon win, their accelerator listing, their public demo, and then match it to a school and a graduation year, not the other way around. The reverse direction (start with school, filter by class year, hope an artifact exists) hits the dead query problem above.

Where to actually source these people

Stop pretending the channels haven't changed. The high-signal surfaces for finding founder engineers from the top engineering school graduates pool in 2026 are not LinkedIn alumni filters. They are:

Student and new-grad accelerators

Y Combinator W26, Breakthrough Ventures' inaugural cohort, UC Berkeley Free Ventures, MIT Sandbox Innovation Fund, Stanford StartX, Stanford LaunchPad, Stanford Cardinal Ventures. Each publishes some form of public batch list, demo day deck, or company directory. Each is a more accurate filter for "top-20 school + actually shipping" than any campus job board.

Hackathon and demo day rosters

Stanford TreeHacks, MIT's hack rosters, Berkeley's Cal Hacks, HackMIT. The winners list from any of these for 2024 and 2025 is a higher-precision filter than "CS, 3.7+ GPA, internship at FAANG."

GitHub commit graphs filtered by graduation cohort

This is the hard one to do manually, which is the point. You want people whose first heavy public-commit period started 2021 to 2023 (sophomore through senior year at a top program) and whose recent activity is on a single repo that looks like a product, not coursework. Doing this across thousands of profiles by hand is a non-starter. Doing it via plain-English search across the same three surfaces (GitHub, LinkedIn, the open web) is the reason a tool like Refolk exists.

"Building" headline searches, semantically

Not "founder" or "CEO" (too noisy, too senior). The actual headline language for this cohort is "Building X," "Working on Y," "Prev Z." A semantic search will catch these. A keyword search will not.

The cohort you can still hire

Here is the optimistic read, because the SignalFire report is not all doom for hiring teams. About half of "top CS grads" are still pursuing traditional employment. The number who founded something is 2x what it was in 2022, but most graduates from the top 20 schools still want a job. They are just facing a 45% lower hit rate at the Tech Majors than the 2022 class did. That is your inbound.

For an early-stage startup or a series B engineering org, this is leverage. A grad who would have gone to Meta in 2022 will now seriously consider a 15-person AI infrastructure startup, because the Meta seat is closed. The same is true at the senior end: the Super IC squeeze means staff engineers are also more available to non-Tech-Major employers than they have been since 2019.

The strategic move is to source both halves of the redirect. The founder half (acqui-hire conversations, "are you sure you don't want to join us instead" outreach, advisory-to-employee paths) and the still-applying half (better, faster, less Leetcode-heavy loops than the Tech Majors are running). Refolk handles both queries in the same plain-English interface, which matters because the candidates often move between the two states over six months.

The close

SignalFire buried a sentence in the report that is worth reading twice: by cutting entry-level hiring to save short-term costs, Big Tech may be incubating its next generation of competitors outside its own walls. Cursor, Cognition, Mercor, and Axiom Math are not coincidences. They are what happens when the top of the funnel gets sealed and the people who would have filled it have access to capital, compute, and AI-native tooling that none of their managers had at 22.

If you are hiring in 2026, the SignalFire State of Talent 2026 numbers are not a story about a shrinking pool. They are a story about a pool that moved. Update your sourcing layer accordingly.

FAQ

Does the SignalFire 65% drop mean new-grad CS hiring is over?

No. It means entry-level hiring at the 12 named Tech Majors is down 65% since 2019, and at early-stage startups it is down 75%. Engineering as a share of new hires at those Tech Majors actually went up (from 46% to 55%). The seats that exist are skewing more senior because laid-off 10+ year ICs are competing for them, but engineering hiring overall is the most resilient category in the report.

How do I actually find new-grad founders if they don't put "Class of 2025" in their LinkedIn headline?

Stop relying on headline keyword search. The accurate signals live across GitHub commit graphs, accelerator batch listings (YC, Breakthrough Ventures, Free Ventures, MIT Sandbox, StartX), hackathon rosters, and personal sites. A semantic search across those surfaces (the use case Refolk is built for) returns the cohort. A boolean keyword search on a single network does not.

Are top engineering school graduates worth the recruiting effort if half of them are starting companies?

Yes, for two reasons. First, about half are still pursuing employment and they are now 45% less likely to land at a Tech Major than the 2022 class was, which means your inbound conversion is structurally higher than it was three years ago. Second, the founder half is reachable later: many early-stage startups fail or pivot within 18 months, and that cohort becomes some of the strongest mid-level engineering talent on the market.

What's the single biggest sourcing mistake teams are making with this cohort?

Treating "no traditional resume markers" as a negative signal. A 23-year-old with one month at "Founder, stealth," a heavy GitHub graph, and a TreeHacks 2025 demo is not a weak candidate with a thin LinkedIn. They are the exact archetype SignalFire's data describes. Recruiters who screen them out because the ATS rejects "founder" as job history are filtering for the wrong column.