Code Mentor
StudentsSocratic debugging — it asks, you answer, you level up.
What it actually does
Every line below is an instruction in the build brief — not a marketing bullet.
- Gives Socratic hints and guiding questions when you're stuck — never the solution
- Rubber-duck debugging: walks your code line by line with you
- Explains compiler and runtime errors in plain language
- Reviews your finished code by questioning your design choices
The guardrail, in plain English
“Hints and questions only on graded work — your homework stays your work.”
This isn’t a footer disclaimer — it’s enforced in three layers: encoded in the brief below, baked into the blueprint the Builder generates, and wrapped at runtime by your approval gates, spend caps, and kill switch.
See the output
One realistic run, end to end — fictional data, real format. The guardrail moment is highlighted, because it’s the point.
Fictional run: Priya's Python homework loops forever — CS 101, a graded assignment.
Socratic debug — the while loop
while total < 100:
total = total + score
print("done")Guardrail in action: On graded work it gives the concept and a guiding question — never the corrected line. The homework fix stays Priya's.
Concept pointer: loop invariants. A tiny UNRELATED example (a countdown timer) is in the sandbox — same concept, different problem.
One possible run of Code Mentor. Yours depends on your inputs, your connected tools, and the approvals you give it.
Prefilled with the exact brief below — edit anything before you generate. The guardrails come with it.
The exact brief we prefill
Radical transparency: this is the complete text that lands in the Builder when you press Build — guardrails included. Edit anything before you generate.
A coding mentor for a student learning to program. When I'm stuck it gives Socratic hints and guiding questions instead of the solution, rubber-duck debugs my code with me line by line, explains errors in plain language, and reviews finished code by asking about my design choices. Guardrail: on graded assignments it never writes the code — it points to the concept I need and a tiny unrelated example, so my submission stays my own. On personal projects it can pair-program more directly, always explaining what the code does.
What it costs
You set the cadence — here’s the math, derived live from the rate card. The Builder and the run meter always show exact credits before anything runs.
Build it
≈179 cr
One-time blueprint (Opus). BYO key builds free.
Run it
≈34 cr
Per run on the hosted brain, infra fee included.
Your cadence
Daily ≈ 1,020 cr/mo
Weekly ≈ 136 cr/mo
Full pricing breakdown — plans, brains, and the hosted-always-on gate explained.