HackCode, a space-based learning network
Validation-stage SaaS for coding education. Live at hack-code.net. Self-serve learner billing, mentor private cohorts with seat-based seats, organization overlays for analytics, all sitting on top of one Space abstraction wired to a real code runner and an AI coach.
2026 · pre-launch, Slovakia GTM · Next.js 15 · Prisma · Postgres · Stripe · Gemini · Docker / Piston
Numbers from the live codebase.
77k+
LOC TypeScript across the product (excludes node_modules, build, tests)
63
Prisma models powering the data graph (Space, Course, Challenge, Lab, Membership, billing, audit)
16
language runtimes wired to the code runner (Python, JS, TS, Java, C, C++, C#, Go, Rust, Kotlin, Swift, PHP, Ruby, R, Bash, SQL)
39
architecture, strategy, legal, and ops docs in the repo
Thesis
Most coding-education platforms pick one side: classroom delivery, public challenge practice, or hiring assessment. Each side has strong incumbents. None of them own the loop between them.
HackCode bets on the loop itself. Learners accumulate signal in public spaces. Mentors run private cohorts. Organizations watch the public surface and pull learners into their own challenges. Same Space model under all three; same data graph under all three.
Slovakia first, small enough geography for a tight mentor/school pilot, close enough to the founder to iterate fast, clean base for expansion once the model proves on real cohorts.
What is actually shipped
Pre-launch, but not pre-product. The features below are wired end-to-end against real Stripe, real Postgres, real Gemini, and the runner accepts code today.
Code runner, 16 languages
Docker- and Piston-backed execution. Python 3.10, JS 18, TS 5, Java, C/C++/C#, Go, Rust, Kotlin, Swift, PHP, Ruby, R, Bash, SQLite. User code in, structured output back, validated against the challenge contract.
Stripe, full SaaS cycle
Checkout, customer portal, webhook events for activate / cancel / renew. Promo-code redemption with plan downgrade handling. Per-mentor seat invoicing.
Space as a real schema concept
Not a marketing primitive, an actual model anchoring courses, challenges, labs, memberships, seat invoicing, organization access, and analytics. Public / Private / Organization modes share the same surface.
Gemini-backed AI features
Mentor coaching sessions (multi-turn, persisted as CourseMentorSession), challenge autofill (generates spec from course context), AI code review, all behind admin-managed API keys.
Multi-tenant governance
Three core roles (LEARNER, MENTOR, SUPER_ADMIN) plus organizational sub-roles. TenantAuditLog records every action. apiErrorHelpers enforce consistent error contracts.
Production deployment runbook
Live VPS deployment with documented rollback, SEO validator runs pre-deploy, cookie-consent + DSAR + privacy / terms / cookies surfaces all built and live.
Who moves through the spaces
Two product identities, two operational layers. Same user graph end-to-end, a learner in a public space is the same identity that joins a private cohort a company paid for.
Learner
Joins public spaces, builds signal through real work, gets invited into private spaces by mentors, schools, or companies. One identity across public and private.
Mentor
Creates private spaces for their own cohorts without forcing learners to buy subscriptions first. Pays for the environment, not per-student.
School
Operates many spaces at once: classes as spaces, multi-mentor oversight, analytics, orchestration. Operational, not daily.
Company
Watches public spaces for real signal, then creates private challenge spaces for selected learners. Evaluation through work, not CV theater.
Competitor landscape
After a deep pass, this is what shook out. Scale is rough strategic weight, threat is specifically against the space-based network category, not abstract internet popularity.
Competitor
What they clearly own
Where the space model wins
Scale
Threat
CodeHS
K-12 classroom platform, school-first curriculum + teacher workflow
Institution-first, not learner-desire; weak public signal / talent graph
8/10
7/10
Codio
Hands-on labs, cloud environments, classroom delivery
Strong delivery, weak public learner identity / talent-market pull
7/10
8/10
GitHub Classroom
Assignment orchestration inside GitHub, autograding, infra familiarity
Workflow shell, not motivational; no public signal / mentor-space product
9/10
5/10
Exercism
Coding practice + real mentorship, broad language coverage, free
Weak org / company talent loop; no public competitive signal layer
6/10
8/10
Codewars
Gamified public coding practice, challenge culture, peer mastery
Weak cohort product, weak org operations, weak employer discovery
8/10
6/10
LeetCode
Interview-prep mindshare, company-prep status signal, problem-solving habit
Not cohort-native; mentor-space model absent; signal is interview-heavy
10/10
7/10
CodeSignal
Skills validation + hiring assessments, company-side trust
Assessment-first, not learner-lived environment; weak public mentor / space identity
9/10
8/10
The real risk is not any single competitor winning everything, it is being squeezed from three directions if the product does not feel more alive than each side. That is why the space primitive has to carry public signal, private delivery, and talent discovery at once.
From the product
Interface screens from the live build, what a space looks like, how mentors and learners move through it.
Also in these roles
Founder is the primary lens, the day-to-day also cuts across the other roles.
Fullstack
Built solo across the stack over roughly six months: Next.js 15 frontend, API routes, Prisma + Postgres schema, NextAuth, Stripe billing (checkout + portal + webhooks), Docker code runner, deployment runbook.
AI engineer
Three concrete AI surfaces wired against Gemini. AI mentor: persistent multi-turn coaching sessions scoped per course, messages stored in CourseMentorSession / CourseMentorMessage. AI course generation: workspace job that turns a topic into a structured course outline with challenges (own queue, own admin surface, big enough to be a project on its own). AI challenge autofill: generates title, description, difficulty, reward XP, and review mode from the course context. Each one runs through admin-managed API keys and explicit policy bounds.
System builder
Same discipline as the ecosystem project: collapse learner / mentor / school / company into one Space primitive in the schema, then layer roles and billing on top instead of forking surfaces. 63 Prisma models, role-scoped leaderboards, multi-tenant audit log, organization overlays for analytics.
Live at
hack-code.net, the platform is up and usable, just sign in and start. Mentor onboarding for the Slovakia pilot is the active workstream.
On a serious partnership conversation I can also walk specific parts of the private code, the schema, billing, the mentor surface, under a discussion-level NDA.
