Show HN: Fata – Spaced repetition to fight skill rot from AI coding
TL;DR Highlight
A spaced-repetition app with daily 10-question quizzes to keep coding skills sharp in the AI era
Who Should Read
Developers who hand most code to agents and worry their hands-on skills are fading
Core Mechanics
- fata is a learning app for developers
- The premise is that you must know frontend, backend, and design basics to direct a coding agent.
- It runs on iOS, Android, and web, and the basics are described as free.
- There is a daily 10-question quiz
- Daily fata needs no account.
- You pick a language and difficulty and can share results.
- Reviews use spaced repetition
- The Core plan lists AI-powered spaced-repetition reviews.
- Offline use, progress tracking, and challenges are also in Core.
- Courses cover SOLID, React, Rust, TypeScript
- Examples shown are SOLID in Python, React Core, Rust for Polyglots, and TypeScript Foundations.
- It also lists topics like CSS, Go, Ruby, Dart, and Architecture.
- Advanced content is in paid Pro
- Pro adds advanced concepts and career-linked pathways.
- The Organization plan offers team management, custom assessments, and reporting.
Community Reactions
- Skill rot is overblown, one commenter said
- After years away from C++, they came back and picked it up quickly.
- They said skills fall behind rather than rot, and the real issue is mental resistance to doing it the hard way.
- Low-level experience stops accumulating, per a comment
- Directing agents lets them build bigger projects but without deep hands-on use of each technology.
- They compared it to shovel versus excavator: using the excavator doesn't erase shoveling skill.
- Pushback on price and differentiation
- One commenter found 13 dollars a month for content that is mostly LLM-generated too much.
- They also said it's unclear how it differs from similar apps.
- Criticism of the copy and claims
- One commenter found 'fata builds' confusing because it sounds like the app does the work.
- Another doubted the '10,000+ developers' claim for a just-launched product.
How to Apply
- Rewrite one agent-written function by hand
- Spend 30 minutes today deleting a small agent-written function and writing it yourself.
- Compare with the agent's version and note what you missed.
- Check your skills with the daily quiz
- Take Daily fata at fata.dev in one language, no account needed.
- Note the topics you missed and retake them the next day.
- Add spaced repetition to team learning
- Turn code-review mistakes and concepts into question cards and review them weekly.
- Use fata or your own cards; either works.
Try It
# Paste into Claude or ChatGPT (simple prompt version of the app's spaced-repetition idea)
You are my coding skill-maintenance quiz coach.
My language: {{language}}
Task I recently delegated to an agent: {{task description}}
Rules:
1. Ask 10 questions one at a time on concepts used in this task.
2. After each answer, say if it's correct and explain briefly.
3. Add wrong answers to a 'review list'.
4. After 10 questions, schedule the review list for the next day, 3 days, and 7 days later.
Ask the first question.Terminology
Related Papers
The Instruction Tax on Factual Recall: Base Models Can Beat Their Instruction-Tuned Counterparts
Few-shot completion with repeated sampling on a base model beat chat: F1 0.684 vs 0.568
AI is a tool and using it well takes skill
A rebuttal to Hickey: LLMs are tools, and skill in directing and checking them decides the outcome
Reflective Prompt Optimization of a Deep-Research Agent: A Topic-Disjoint, Multi-Seed Protocol for Measuring Transferable Gains and Prompt-Level Overfitting
Reflective prompt optimization raised held-out citation reward from 0.702 to 0.781 on a deep-research agent.
GPT‑6 and Intelligent UI for everyone
GPT-6 reportedly adds generated UI to chat; with no article text, only comment reactions are available.
Base Models Can Reason By Taking a Cue From Training Data
Prefilling just two opening tokens lifted Olmo-3-7B's MATH-500 accuracy from 42% to 78%.
Labels Override Definitions in Jev-Style Typed Decision Models
Renaming options to A/B raised routing accuracy by +0.1511 over meaningful labels.
How Much Prompt Is Enough? A Blackbox Minimization of Few-Shots in LLMs
Delta debugging trimmed few-shot prompts by 65.3% on average with unchanged propositional output
Language Models Are "Insecure" Reporters
Adding "Be honest" raised GPT-5.5's flagging of a hidden negative result from 2/200 to 190/200.