On June 11, Show HN: Fata – Spaced repetition to fight skill rot from AI coding hit the front page of Hacker News, with 75 points and 44 comments. Fata itself is a kind of Duolingo for code: 5–10 minutes a day of spaced repetition on Rust, CSS, React, Python, TypeScript, and Architecture, runnable in the browser with no signup required.

But the thing that actually caught my attention is not Fata. It is how the author describes himself.

Hi HN, I’m Djoumé. I’ve been a developer for over 20 years, and like a lot of you I’ve been coding almost exclusively through an agent in the past few months. It’s been amazing to vibe code prototypes in any stack, but when it comes to building something reliable/scalable, I couldn’t effectively guide the agent unless I knew the technology. And the scariest part is that I’m seeing a lot of my technical skills decreasing due to AI coding.

Read that paragraph on its own. A developer who has been writing code for twenty years noticed, after a few months of letting an agent do most of the typing, that his own technical skills are getting worse. That is the real reason Fata exists.

1. Fata is not solving “the beginner can’t learn”, it’s solving “the veteran is forgetting”

Spaced repetition (the Anki / RemNote family) is something we all know. It used to serve students and exam-preppers: the cards you keep forgetting get scheduled more often. What Djoumé built is the same mechanism, pointed at a different audience: seasoned developers who can feel their edge dulling.

Fata’s homepage does not pitch “learn from zero.” It pitches “stay sharp.” The questions inside a course — Rust lifetime annotations, React reconciliation timing, TypeScript conditional types — are exactly the kind of thing you could write with your eyes closed two years ago and now need to look up to get right.

Side by side with traditional programming education, the difference looks like this:

Dimension Traditional course (Codecademy / Frontend Masters) Fata
Audience Career-switchers, beginners Already a developer, worried about going soft
Rhythm Sequential chapters 5–10 minutes a day, spaced repetition
Depth Hello world to production Assumes you know it; targets what you forget
Assessment Capstone projects Repeated recall, no decay
Pricing Hundreds of dollars, one-time $13 / month

Fata’s real pitch is the acknowledgement of an awkward fact: AI makes us write code faster, but “able to write code” and “able to guide an agent into writing reliable code” are not the same skill.

2. How Fata builds itself: the meta-story

In the same Show HN thread, Djoumé also drops the engineering details. Worth quoting in full:

The first courses were painfully written by hand, but most content is now AI-generated. It takes about 3000 LLM calls to generate a course, and every code samples goes through compilation, linting, unit testing, AI and a final manual review.

3,000 LLM calls per course. Every code sample passes five gates: compilation, lint, unit tests, AI self-review, and a final human review.

I paused on this one for a while. The entire content production pipeline for this product is AI writes → AI plus tools verify → human reviews. That is the same template we keep seeing on the radar for “AI-native teams.” Fata is, itself, the template for the problem it is trying to solve.

But there is an extra twist here. Fata built this pipeline for one purpose: to keep other people from depending on AI to write code. It batch-produces courses with AI while selling users on staying manually capable. This is not a contradiction. It is the core of the business model: what Fata sells is not the cure for skill rot, it is the early-intervention for skill rot.

3. What the HN thread is really arguing about

75 points and 44 comments is busier than the average Show HN. Reading through, the conversation is barely about whether Fata is good. It is about three new anxieties about what a developer is supposed to be.

3.1 The “how does the next generation grow up” anxiety

Djoumé drops one of the heavier lines in the post itself:

Reflecting on my journey, I also worry about how the new “AI native” generation of software developer is going to acquire technical depth.

AmblingAvocado pushes back with what sounds reasonable:

I stopped writing C++ professionally in the early 2010’s. I came back to it in the early 2020’s and it was extremely easy to pick up. Less like riding a bike, more like driving a car.

Then danielmarkbruce answers sharper:

No one said anything about learning everything before doing anything, and nobody suggested they were told to do it or how they did it. And, bad engineers can become good ones. All the good ones were terrible at some point.

Read those two replies side by side and the tension shows up clearly: the long, painful apprenticeship that produced technical depth in the previous generation is being skipped by agentic coding. Fata is not selling code ability. Fata is selling a retrofitted patch for skipped apprenticeships.

3.2 The “my own hands are getting stiff” anxiety

bluefirebrand calls out Fata’s homepage copy as confusing:

“Fata builds” sounds like it does the work, not like it trains you how to do the work. With how many “AI does the work for you” promises out there, it’s confusing terminology.

The comment is about a wording bug, but it exposes something deeper: when “AI does it for you” is the default story, even a product that helps you do it yourself has to fight for its positioning. Fata has to convince users to “practice so they don’t get replaced,” which is a much harder sell than “study so you get promoted.”

3.3 The “am I paying to practice AI-generated content” anxiety

The most interesting objection comes from Falimonda:

If you need to use an agent then you might as well just work SRS into your harness using something like srs.voxos.ai

The point: if you already run an agent, just bake spaced repetition into your Claude Code or Codex harness. No need to pay $13 a month for a separate website.

This comment puts a question Fata has to answer on the table: in agentic coding, where should “practice” actually happen? Fata picks “10 minutes a day in a separate browser tab.” voxos.ai picks “live inside the writing flow.” Two routes, two bets on the future — Fata bets “developers will be pushed out of the agent flow and need dedicated practice time.” voxos.ai bets “the agent harness will absorb everything.”

4. Why this surfaced in June 2026

Fata is not an isolated product launch. Three timelines collided here.

First, agentic coding crossed into the “main output” stage. The current generation of Anthropic Claude Code and OpenAI Codex agents, by mid-2026, moved from “use occasionally” to “use daily.” Djoumé says “past few months,” not “past year” — the skill-rot feeling becomes noticeable only after 6–9 months of high-intensity agent use, not at the start.

Second, spaced repetition as a mental model returned to developer tools. Between 2024 and 2025, the Anki / RemNote style saw renewed attention as LLMs were applied to flashcard generation. Installations of the Obsidian Spaced Repetition plugin roughly doubled in 2025. The vocabulary of “review, don’t cram” re-entered the developer stack, which is why Fata had a ready audience.

Third, Fata’s stack says something about its bet. The app is offline-first mobile, built with Capacitor, RxDB, and Firebase. That choice means Fata is betting that developers will spend fragmented mobile minutes per day practising code. The growth of GitHub Mobile in 2025 is what underwrites that bet.

5. What Fata tells me, beyond what Fata meant to say

The biggest thing I took from this Show HN is not whether Fata is worth using. It is that Fata named a new anxiety.

“Skill rot” is not a phrase that was common in 2024 or earlier. Search the English-language tech press and the contexts it appeared in were:

  • Soldiers losing combat readiness after long deployments
  • Pianists losing finger facility after years off the bench
  • Senior physicians losing diagnostic accuracy after leaving the clinic

In June 2026, that phrase got attached, for the first time at scale, to software developers. That linguistic drift is itself the signal: the developer community is starting to admit that “can use AI to write code” and “can write code” are diverging, and the gap is widening.

Fata’s answer is “use spaced repetition to fight back.” But the question is much bigger than one product. It is unfolding in three directions at once:

  • Career direction. Do AI-native teams still need senior engineers? The answer keeps tilting to “yes,” but the reason keeps shifting from “writes complex code” to “judges whether the agent wrote the right code.”
  • Learning direction. Do new developers still need the “bad-to-good” apprenticeship? Djoumé is asking. AmblingAvocado is disagreeing. The industry has no consensus.
  • Product direction. Is the next generation of developer tools about “make the agent more capable” or “keep the human from decaying”? voxos.ai and Fata both bet on the latter.

None of these gets answered by one product. But Fata at least put the question on the table.

6. If your hands are getting stiff inside agentic coding

A few responses I have seen, tried, or read in the comments:

  • Reserve 20% of your time for “write without the agent.” Djoumé’s 5–10 minutes is short, but the principle is right: keep a daily window of pure hand-writing.
  • Use Anki. You do not need Fata. Anki is open source, and you can author your own cards. But for most developers, Fata’s curated courses are a much lower barrier than authoring Anki decks from scratch.
  • Bake spaced repetition into your agent harness. If you already live inside Claude Code or Codex, follow voxos.ai’s lead: write a skill or subagent that forces you to recall at key decision points. This is the more aggressive route.
  • Stop treating “my hands are getting stiff” as a personal failing. It is a structural side effect of the agentic coding era. Admitting that is more honest than pretending you are still sharp.

Fata itself may or may not survive. But the fact that “skill rot” is now an accepted phrase for developers is not going back.

References