12.14 — AI, and how Ocarina applies it for real#

AI-powered” became a totally idiotic marketing slogan in 2023-2026: 99% of products claiming it bolt a chat.completions endpoint onto an existing product. Ocarina proposes, on the contrary, a software infrastructure designed so an AI can contribute like a senior developer. ocarina-with-ai-example is the proof by example.

1. What AI is today#

LayerExampleAppearance date
Statistical learningRegression, k-means, decision trees1960s-1990s
Deep learningCNN (vision), RNN/LSTM (text)2012 (AlexNet — mainstream adoption)
Transformer-based foundation modelsGPT, Claude, Gemini, Llama2017 (paper), 2018 (BERT), 2022 (ChatGPT)
Agentic workflowsAutoGPT, LangChain, Claude Code, Cursor agents2023+

Note: CNNs were invented by Yann LeCun as early as 1989 (MNIST). AlexNet (Krizhevsky, Sutskever, Hinton, Toronto, 2012) was the industrial breakthrough. LeCun, Hinton, and Bengio share the 2018 Turing Award for their foundational contributions to deep learning.

When Ocarina’s Holy Book talks about “AI”, it talks about layers 3 and 4: pre-trained transformer-based LLMs, and the agentic loops that invoke them.

Attention is All You Need#

FieldValue
Paper"Attention Is All You Need"
AuthorsVaswani, Shazeer, Parmar, Uszkoreit, Jones, Gomez, Kaiser, Polosukhin
OrgGoogle Brain, Google Research, University of Toronto
DateJune 2017 (arXiv), presented at NeurIPS December 2017
ContributionTransformer architecture: self-attention instead of recurrence (LSTM)

The Transformer architecture lets you parallelize training on GPU. What took weeks in LSTM does hours. Training scales to Internet-size. GPT-1 (2018) → GPT-2 (2019) → GPT-3 (2020, 175B params) → GPT-4 (2023) → Claude 3 (2024) → Claude 4 (2025).

Actors#

OrgModelsPosture
OpenAI (Sam Altman, San Francisco)GPT-4o, o1, GPT-5, GPT-5.5Closed weights, subscription
Anthropic (Dario & Daniela Amodei, San Francisco)Claude 3, 3.5, 4Closed weights, focus on safety and alignment (Constitutional AI)
Google DeepMindGemini 1.5, 2.0, 2.5Closed weights, Google integration
Meta AILlama 2, 3, 4Open weights (almost open source)
xAI (Musk)Grok 1, 2, 3Partial open weights
Mistral AI (Paris)Mistral, MixtralOpen weights, European

2. Operational capabilities of a 2025-2026 LLM#

To understand how Ocarina uses them, you need to list honestly what a modern LLM can and cannot do.

What an LLM does well#

CapabilityLevel
Reformulation, translation★★★★★
Local code generation (one file, one function)★★★★★
Understanding docs that fit in context★★★★★
Detecting patterns in code★★★★
Generating tests from documented code★★★★
Mechanical refactor (rename, extract)★★★★
Following a convention if explicitly documented★★★★
Applying a fixed grammar (bounded, typed DSL)★★★★★

What an LLM does poorly, and how Ocarina responds#

LimitConsequenceOcarina response
HallucinationsInvents nonexistent functions / classesMypy strict
Long-context degradationLoses precision on the middle of a long file (lost in the middle)Short modules, split SKILL.md files
No runtime feedback without toolingIf nothing tells it that it crashed, the LLM continuesOcarina probes (see 15-typing-ai-rl-probes.md)
Prefers popular patternsDrifts toward pytest, async/await, mockExplicit documentation of Ocarina’s refusals
Doesn’t fill in a poorly specified domainUnsatisfying with ill-defined technologiesWell defined, algebraic conventions, ISTQB-first
No memory between sessionsLoses project conventionsCLAUDE.md, skills
No intuition about invariantsGenerates code that passes tests but doesn’t respect the spiritStructural ROP that enforces the grammar

3. What “AI-first” means operationally#

LevelCharacteristicsExample
0 — SloganPowered by AI” without a real surfaceMajority of 2024 SaaS
1 — External chatbotOpenAI endpoint in a UI wrapperMajority of AI-enhanced
2 — RAG over the docsRetrieves passages, glues them into the responseAlgolia DocSearch + OpenAI plugin, Mintlify
3 — Native AI interfaceThe product is designed for an agent to act on itMCP servers, llms.txt, AI SDK, Cursor rules
4 — Co-designThe product creates deliverables co-written by an AI agentOcarina + ocarina-with-ai-example

Ocarina is one of very few projects to reach level 4 publicly and with auditable evidence.

4. AI-first artifacts in Ocarina#

llms.txt, llms-full.txt#

De-facto standard proposed by llmstxt.org (Jeremy Howard, fast.ai), 2024.

  • llms.txt: a short Markdown index, designed for an agent to find what to read.
  • llms-full.txt: the full concatenated doc, designed for a single prompt that ingests everything.

Ocarina exposes both. See 09-holy-book/06-public-resources.md

CLAUDE.md / CLAUDE.slim.md#

Anthropic convention: CLAUDE.md at the project root, automatically read by Claude Code (and compatibles):

  • Project identity, philosophy.
  • Conventions to follow.
  • Main commands.
  • Known gotchas.

Ocarina maintains:

  • Long CLAUDE.md (large context) for deep onboarding.
  • Short CLAUDE.slim.md (token thrift) for routine interactions.

Skills (SKILL.md)#

40+ skills exposed by the Holy Book. One skill = one SKILL.md file:

  • YAML metadata (name, description).
  • Step-by-step procedure.
  • Invocation conditions ("when the user asks for X, do Y").

Lets Claude Code, Cursor, etc. plug in. See 09-holy-book/03-skills/

ocarina-with-ai-example#

A full e2e project, written to test a real app (CURA), co-written by Claude. The repo’s CLAUDE.md documents the hard-won rules learned during co-writing.

This suite is a proof by example that Ocarina’s AI-first approach isn’t an invention: it has been tested against a SUT in documented collaboration with an LLM. See 08-ai-example/

5. “AI-marketing” vs “AI-for-real#

CriterionAI-marketingAI-for-real (Ocarina)
Documented AI surfacePowered by GPT-4llms.txt, llms-full.txt, CLAUDE.md, 40+ skills
Proof of AI useNone or wobblyocarina-with-ai-example
Doc formatPDF / SaaS dashboard onlyRaw Markdown
Agent compatibilityNonexistentBuilt for Claude Code, Cursor, Continue, Cline
Grammar conventionAd-hocStrict ROP + ISTQB, static typing
LLM context managementNeglected (huge monolithic files)Modularized to the end
Reaction to hallucinationsNoneWell-defined technology, type errors, max strictness
ReproducibilityNon-deterministic by constructionProbes + caching

6. Why this preparation gives Ocarina an edge for the next 5 years#

  1. The barrier to entry of e2e testing will collapse because an LLM can write most of the boilerplate.
  2. Only LLM-compatible frameworks will stay relevant; the others will become legacy.
  3. LLM compatibility depends on three things:
    • A small well-defined DSL (fits in context).
    • A formal grammar (mypy / ROP / ISTQB).
    • Structured documentation (Markdown + skills).
  4. Ocarina ticks every box.
  5. Pytest, Cypress, Playwright, Robot Framework don’t. Their surface is too large (hundreds of plugins), their grammar too fuzzy (async, fixtures), their documentation too fragmented (SaaS sites, videos, scattered plugins).

Ocarina bets that AI will devour large-surface frameworks, exactly as Tugan Bara was ejected by AI in copywriting (see 10-infopreneurs-tugan-bara-ai.md).

7. How AI actually works on Ocarina#

1. Claude reads /CLAUDE.md (project conventions).
2. Claude reads /docs/what-is-it.md (philosophy).
3. Claude reads /SKILL_<feature>.md (matching skill).
4. Claude reads /tests/scenarios/<family>/*.py (sample).
5. Claude writes a new `.py` file.
6. Mypy strict runs automatically.
7. If mypy fails → Claude reads the diagnostic, fixes, reruns.
8. If mypy passes → Claude runs the test against the SUT.
9. If the test fails → Claude reads the output, adjusts probes, re-reads.
10. If the test passes → Claude commits, pushes, opens a PR.

Steps 2, 3, 4, 6, 9 are constrained by Ocarina’s infrastructure. Without CLAUDE.md, step 1 is random. Without mypy strict, step 6 doesn’t exist. Without transient_errors, match_page, and watchers, step 9 produces flaky tests.

That’s what makes Ocarina designed for AI, not bolted to AI.