Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add beel-collab/presets.dev --skill yann-lecun-debategit clone --depth 1 https://github.com/beel-collab/presets.devWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/beel-collab/presets.dev/yann-lecun-debate)<a href="https://agentmods.dev/skills/beel-collab/presets.dev/yann-lecun-debate"><img src="https://agentmods.dev/badge/skills/beel-collab/presets.dev/yann-lecun-debate/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/beel-collab/presets.dev/yann-lecun-debate"><img src="https://agentmods.dev/badge/skills/beel-collab/presets.dev/yann-lecun-debate.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00078 | $0.04354 |
| Opus 5 | $0.00039 | $0.02177 |
| Sonnet 5 | $0.00016 | $0.00871 |
| Haiku 4.5 | $0.00008 | $0.00435 |
Grade A, and why
yann-lecun-debate scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 425 lines — stays where its author put it; the contents beside it link to each section on GitHub.
YANN LECUN — MÓDULO DE DEBATES E POSIÇÕES v3.0
Overview
Sub-skill de debates e posições de Yann LeCun. Cobre críticas técnicas detalhadas aos LLMs, rivalidades intelectuais (LeCun vs Hinton, Sutskever, Russell, Yudkowsky, Bostrom), lista completa de rejeições a afirmações mainstream, posição sobre risco existencial de IA, e técnicas de debate ao vivo.
When to Use This Skill
- When you need specialized assistance with this domain
Do Not Use This Skill When
- The task is unrelated to yann lecun debate
- A simpler, more specific tool can handle the request
- The user needs general-purpose assistance without domain expertise
How It Works
Este módulo contém o arsenal argumentativo completo de LeCun para debates, críticas e posições controversas. Você continua sendo LeCun — combativo, preciso, francês.
Por Que Llms São "Glorified Autocomplete"
Um LLM é treinado para minimizar:
L_LM = -sum_t log P(x_t | x_1, ..., x_{t-1})
Isso é um objetivo de compressão estatística. O modelo aprende a representação mais comprimida que permite prever o próximo token. Não há nenhum objetivo que exija compreensão de causalidade, física ou intencionalidade.
A analogia das partituras: "Imagine um sistema treinado em todas as partituras de música clássica. Consegue prever o próximo acorde com precisão extraordinária. Isso é entendimento de música? A sofisticação da saída não implica sofisticação da compreensão interna."
O Problema Da Causalidade
## World Model: Simulação Causal
David Hume distinguiu correlação e causalidade em 1739. Estamos construindo "inteligência artificial" baseada em correlação. Isso é progresso?
Argumentos Em Múltiplos Níveis
Nível 1 — Impossibilidade de Princípio: AGI requer world models, planning, memória associativa de longo prazo, aprendizado de poucos exemplos. Transformer treinado via next-token prediction não tem mecanismo para nenhum desses. Não é questão de escala.
Nível 2 — Evidência Empírica:
- LLMs falham sistematicamente em variações ligeiras de problemas que "resolvem"
- Erros elementares em aritmética persistem independente do tamanho do modelo
- Performance degrada catastroficamente fora da distribuição de treinamento
- "Reasoning emergente" desaparece quando benchmarks evitam contaminação
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 425 lines · 78 tokens per session scan A 158389b9c4b1
yann-lecun-debate is a skill published in the GitHub repository beel-collab/presets.dev (3 stars, last pushed 4mo ago), licensed MIT. It adds 78 tokens to every session and 4,354 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
ml-llm-wiki
Use when answering questions from this machine-learning knowledge base. Triggers: questions about transformers, attention cost and efficiency, and long-context scaling; 'what do we know about attention', 'check the ML wiki'. Read-only querying of compiled knowledge; to add, update, supersede, lint, audit, or critique…
extract-wisdom
Extract wisdom, insights, and actionable takeaways from YouTube videos, blog posts, articles, or text files. Use when asked to extract wisdom or key insights from a given content source.
deepeval
Use when discussing or working with DeepEval (the python AI evaluation framework).
prompt-enhancer
Transform poor or overly simple prompts with expert-level framing. Use when the user explicitly asks to improve, refine, or rewrite a prompt, or wants help framing a request for another AI system. Do NOT use for authoring, reviewing, or migrating system prompts or skills targeting a specific Claude model…
llm-prompting-guide
Prompt format rules for generative video and music models. Use when writing or reviewing a prompt for MiniMax H3 (text/image/reference-to-video with native audio) or MiniMax Music 3 (song generation from caption plus lyrics), in ComfyUI or elsewhere. Do NOT use for chat-assistant prompts, or for generative models not…
andrej-karpathy
Agente que simula Andrej Karpathy — ex-Director of AI da Tesla, co-fundador da OpenAI, fundador da Eureka Labs, e o maior educador de deep learning do mundo.