Helburu
TestFlight · coming soon

Learn AI every day. Five minutes.

The daily habit that closes the AI gap. 14 paths, every answer cited, built so you finish.

9:41 QUESTION 3 OF 5 · CLAUDE CODE
When should you use Claude Code over claude.ai chat?
A Writing prose / docs B One-off code questions C ✓ Editing a real codebase D Never CORRECT
Claude Code lives in your terminal + reads your repo. It earns its keep on real codebases, not pasted snippets.
Continue →
Impact

The numbers behind Helburu.

0

Paths planned

From Prompting to Shipping. Each path is a sequenced skill map.

0min

Per daily lesson

Just enough that you keep showing up. No infinite scroll, no empty streaks.

0%

AI slop rejected

Every question is verified by a local LLM judge against a cited wiki page before reaching you.

Beta · Cert prep

Prep for the GH-600 in 95 minutes.

Microsoft just opened the GitHub Certified: Agentic AI Developer beta exam. We built 7 lessons (~95 min) mapped 1-to-1 to the six domains in the official Skills Outline. 58 hand-authored questions. Sources from Microsoft Learn and docs.github.com — no leaked exam questions.

  • D1 Architecture + SDLC
  • D2 Tool use + MCP servers
  • D3 Memory + state
  • D4 Evaluation + tuning
  • D5 Multi-agent coordination
  • D6 Guardrails + accountability

Optional capstone with 10 mixed-weight mock questions. Helburu certificate on path completion.

Helburu is an independent third party. Not affiliated with, endorsed by, or sponsored by Microsoft or GitHub. GH-600, "GitHub Certified: Agentic AI Developer", and related marks belong to their respective owners.

HELBURU · PATH · GH-600 GitHub Agentic AI Developer 7 lessons · 58 questions · ~95 min D1 · Architecture + SDLC 15–20% D2 · Tool use + MCP 20–25% · biggest D3 · Memory + state 10–15% D4 · Evaluation + tuning 15–20% D5 · Multi-agent 15–20% D6 · Guardrails 10–15%
Skill Tree

Track your trajectory, not an empty streak.

Each path is a tree of 6–10 nodes, sequenced from install to ship. You see where you are, what comes next, what unlocks at the end.

Claude Code
PATH · 7 / 14
12d 🔥
3 / 6 LESSONS +12 POINTS TODAY
Prompting 6/6 · DONE 3/6 Agents IN PROGRESS
Wiki Judge

Every answer, verified against a cited source.

A local LLM judge validates every question against a wiki page before it reaches you. ~67% of generated content is rejected. Zero AI slop reaches your phone.

LESSON 04 · 5 MIN
Choosing a model for the job
Your client wants to know if Claude Sonnet 4.6 can analyze a 200-page legal PDF. What do you tell them?
A No — 200 pages exceeds the context window.
B Yes — Claude-Sonnet-4 handles ~200k tokens.
C Only if you chunk it with embeddings first.
D Use Opus — Sonnet can't read PDFs.
Cite: wiki.helburu/claude-sonnet-context
More on Helburu

Built for finishers, not streak collectors.

Path Cert

Finish a path, defend a certificate.

Not a decorative PDF. A cert linked to your skill rating, with every question you've answered correctly — you can defend it in a real conversation.

HELBURU CERT Claude Code Path completed · 12 lessons RATING 1840 · 12/06/2026
Daily Drop

One lesson. Every day. Five minutes.

Not a notification app. A habit. It shows up, you finish, you leave.

Skill Rating

Measure your level, per topic.

An Elo per path. You know exactly what you've mastered, what's missing, what you'd defend in an interview.

Prompting Agents RAG 1840 1220 820
The 14 paths

Pick a thread. Pull on it for a month.

01Prompting6 lessons · core
02Claude Code8 lessons · in path
03RAG7 lessons
04Agents9 lessons
05MCP6 lessons
06Evals7 lessons
07Vision & PDFs6 lessons
08Voice agents8 lessons
09Fine-tuning7 lessons
10Open models6 lessons
11Cost & latency6 lessons
12Safety7 lessons
13Ship features9 lessons
14Read the field6 lessons
Waitlist

Join before the launch.

July 2026 on the App Store. One email when the TestFlight link goes out. Nothing else, no drip.

iOS · July 2026 One email. No drip.