Half your code is now written by an AI that hallucinates the other half's package names.
That's not a joke.
Across 576,000 samples, 19.7% of the packages AI coding models suggested didn't exist — and attackers have started pre-registering those exact hallucinated names, waiting for someone to a package that was never real to begin with.
Meanwhile, public registries catalogued 454,600 new malicious packages in 2025 alone, pushing the cumulative total past 1.2 million.
A modern web app now pulls in 1,200+ dependencies once you count the full transitive tree — for software that, ten years ago, would have shipped with a fraction of that.
That's precisely what Zero Dependency 2026 is about.
Organized by Hackathon Raptors, Zero Dependency is a 72-hour online global hackathon where participants build genuinely useful software — CLIs, parsers, servers, databases, security tools — using nothing but their programming language's standard library.
No frameworks.
No packages.
No supply chain to inherit.
Just your language and your engineering skill.
Why Zero Dependency?
Modern AI coding assistants can generate code fast, and they reach for a dependency by reflex — sometimes one that doesn't exist.
But real engineering isn't judged by how many packages you imported.
It's judged by questions like: Do you actually understand what the library you'd normally import is doing underneath?
Can you build the thing yourself, correctly, from primitives?
Does your code handle the edge cases a package would've handled for you?
Is the implementation idiomatic, or a fight against the standard library?
Could a judge verify your dependency manifest is empty in five seconds?
Zero Dependency is designed around these questions.
Instead of rewarding whoever pulls in the most packages the fastest, the hackathon rewards teams who can prove they didn't need to.
About the Hackathon 🌍 Global Online Hackathon ⏳ 72 Hours 💰 $1,800 Prize Pool 👥 Solo or Teams of up to 4 🎓 Open to students, professionals, researchers, and developers worldwide 💸 Completely Free to Participate Whether you're a backend engineer, a tooling/DX person, a security engineer, a database-curious systems programmer, or just someone who's felt something die inside after -ing a one-line function — Zero Dependency offers a genuinely different kind of challenge.
What You'll Build Every participating team will: Pick one of six tracks.
Define a real problem worth solving.
Design and implement the solution using only their language's standard library.
Test the implementation against real edge cases.
Document every technical decision and trade-off.
Verify the final artifact has zero third-party runtime dependencies.
Submit a working demo and full dependency proof.
Unlike many hackathons, success isn't determined by flashy pitches.
The emphasis is on functionality, honest craft, idiomatic code, and a dependency manifest a judge can verify at a glance.
Six Tracks Track A — Developer Tools & CLI Linters, formatters, task runners, git utilities, file utilities.
The daily-driver tools most people reach for a dozen packages to build.
Track B — Parsers & Data Formats JSON/CSV/Markdown/config parsers, template engines, regex engines, serializers.
The things everyone imports and almost nobody has written.
Track C — Web & Network HTTP servers, routers, clients, static-site servers, DNS tools, raw TCP chat apps.
Built on stdlib networking primitives alone.
Track D — Data & Storage Key-value stores, embedded databases, caches, log-structured stores, search indexes.
The layer most apps rent from a library and never look inside.
Track E — Security & Crypto Utilities Password managers, TOTP/2FA generators, file encryptors, hashers, secrets scanners.
Compose trusted stdlib crypto primitives correctly — never roll your own cipher.
Track F — Open / Wildcard Games, visualizers, interpreters, compression tools, schedulers — anything genuinely useful that a reasonable engineer would assume needs packages.
Justify the build