The Gabriel Petersson playbook — getting hired by not competing

01TL;DR

Gabriel Petersson — Swedish high-school dropout, cofounder of Depict.ai, then Dataland → Midjourney (2023) → OpenAI (Dec 2024, Sora team, Cameos) — has one consistent message about the job market across a year of threads: the standard application track is the only genuinely competitive part of hiring, so opt out of it. Companies don't want credentials; they want to know if you are good at the specific thing they need. Prove that with a personalized demo, hand it to decision-makers directly, remove their risk with a trial offer, and run your career as a handful of deliberate decisions instead of a pile of applications.

He used this exact system to land every role on his way up — including OpenAI, which had rejected him through the normal process a year earlier.

02The thesis: never compete

The opening move of his signature thread (Mar 9, 2025) is the whole philosophy in two sentences:

His model of what hiring actually is: companies just want to know if you're good at what they need. If you can explain that simply and concretely, then "hiring you vs some perfect grade university student is a no brainer — you'll be the less risky option." Note the frame: the demo doesn't just display skill, it changes the risk calculus of the person on the other side of the table. A credential is a probability; a working artifact in your hands is a fact.

He also argues the competition itself is mostly illusory — careers are nothing like the zero-sum arenas people fear (Mar 19, 2025):

In other words: job seekers cluster on the same narrow actions (CV templates, portal applications), never observe the results, and never iterate. Simply acting on feedback puts you ahead of nearly everyone.

An engineering manager quoting the thread back at him confirmed the employer side: "hiring would be 1000× easier if people focused on demos instead of optimizing credentials on CVs that I won't read anyway — because I only interview people who show off their work."

03The demo-first playbook

Build something for them, not a portfolio

Not a generic portfolio project — a small, working product tailored to the target company's mission. His own examples, both still live on his GitHub:

• OpenAI — a working website demo of what he thought OpenAI was building; this got him "all my interviews… over two years ago before moving to sf"
• Midjourney — "when interviewing at midjourney, i spent a weekend making a full canvas image generation tool to make sure they knew exactly what i'm good at. don't wait for them to ask interview questions"
• For a friend applying to AI jobs — he suggested building "a diffusion model that solves mazes": super cool, and instantly legible as AI skill

Optimize for the 5-second read

Demo selection is an attention problem first:

Record a video walkthrough

When he applied the system to Midjourney, he followed the demo with a video walking through the code. As he told Fortune (2026): "I show my understanding, I show that I'm good socially. They can see this person seems reasonable. I tick more boxes than I ever could by any proxy." One artifact — code, taste, communication, judgment — no proxies needed.

The funnel

04Distribution: skip the portal

Bypass the normal application process "generally"

His explicit instruction: "also don't apply through normal application process generally." When asked whether the demo goes to HR, his answer was blunt — go "straight to managers, ceo, ppl with incentive for the company to go well."

Why not HR

HR is structurally incentivized to reject anomaly and accept pedigree — the demo candidate is precisely the anomaly their incentive system punishes. Decision-makers, by contrast, are rewarded for the company doing well, so they're the ones who can act on proof.

Maximize surface area

• Direct outreach: "mostly email ppl, just send them links and a short personalized message" (his suggested script, per Fortune: "I was so excited about your company that I've been having this side project of building an actual website for what you guys are doing")
• Public: "go to events, show everyone your demo. post it on all social media and try to go viral"
• Referral flywheel: "shove the demo in peoples faces — people will immediately get more helpful to intro you to other companies when you remove all doubt about your abilities. it's a win win for all parties"

The Midjourney role is what opened the OpenAI door — a friend connected him to the research team at a company that had already rejected him once. His lesson: retry after you can show more. The rejection wasn't a verdict on his ability, just on the evidence he'd presented.

05Reverse the risk

The demo proves ability; the second move removes commitment anxiety. From the same thread:

Two years later (May 31, 2026 thread) he generalized this into his advice for early-career engineers — the same mechanism, aimed at the job market as a whole:

• Frame short engagements as internships, "try working together for a month," or contract work — this makes entry cheap and hopping frictionless
• The win condition is information symmetry: "huge wins for everyone that all sides have information so you can price yourself in" — you learn what good teams look like; the market learns what you're worth
• Counter-example he warns against: great engineers who spent years at companies that were not stepping stones ("2.5 years at a startup" at $80k). Landing at a frontier lab or hot startup early is real but "extremely rare" — don't bet on it, buy datapoints instead

06Interviews: pitch, don't perform

Take control

The interview becomes a demo session instead of a LeetCode/reg trivia gauntlet — which is exactly the environment where the credential gap stops mattering.

Speak in concrete terms only

After interviewing 50+ people himself (Jul 2026 thread), his sharpest complaint:

The inverse also holds — his diagnosis of people who can't get a job (Aug 29, 2025): they fail on "how to show you are good," usually by taking CV advice and then "spew[ing] out noise and reverse signal… and talk about how much they like hiking."

07The market as a decision problem

The most Tweet-quoted of his career threads (Mar 13, 2025):

The three neglected decisions

His Aug 29, 2025 thread: "90% of big career decision mistakes happens in these, and usually people spend ~0 seconds of their life thinking about them":

1. when to leave your current company — people "never consider leaving except if the job is miserable, and that's way too late"
2. make sure others understand how good you are — the demonstration problem; this is where most failed searches actually break
3. talk to enough companies to make a 1%-percentile decision — "then accept the first offer they get" is the default failure mode

The inaction asymmetry

Underneath the tactical advice sits a psychology thread (Mar 5, 2025) about why people stay stuck: "70% of people are in permanent slight suffering because they are allergic to making ANY mentally tough decision when there is also an option to do nothing." Friends will help you cope, not decide ("the moment they see emotion in you, they'll stop helping and start helping you cope"). His closing line is the whole worldview compressed:

Related warnings from the same year: don't be semi-ambitious — "you might end up working 60h weeks for 30% higher salary in a middle management position with people you don't like… and they stick around forever." And don't confuse motion with decisions — "the feeling of productivity doesn't matter."

08Credentials, university, and AI

The playbook assumes the credential race is unwinnable and unnecessary — and he says so directly:

• Universities lost the monopoly: "universities don't have monopoly on foundational knowledge anymore" — he taught himself the math behind diffusion models from zero probability background using ChatGPT, drilling top-down from real projects ("you start with a problem, you recursively go down")
• Results are the currency: "companies just want to make money. you show them how to make money, that you can code, and they'll hire you" — and his viral line: "if you're a smart person who can use ChatGPT, you can get a job tomorrow"
• AI is a lever, not a threat: at a moment when entry-level roles are being squeezed by AI, his answer is to adopt the tools faster than the market and let artifacts — not diplomas — carry the proof
• On university itself (25-tweet thread, Mar 20, 2025): a highly individual decision — he dropped out of high school and sees university as worth it mostly for the people and fun, "if you don't genuinely want to become a math professor or something"

09Caveats & counterpoints

10Source trail