https://beltran12138.github.io Product & research at the intersection of AI agents and crypto markets. en https://beltran12138.github.io/writing/the-same-you-scored-differently https://beltran12138.github.io/writing/the-same-you-scored-differently Changing two backend settings tripled a model's ARC-AGI-3 score without touching a single weight. Capability is a joint product of the model and the harness — and the machine now reading your résumé prefers sentences it wrote itself, 67 to 82% of the time. Thu, 13 Aug 2026 00:00:00 GMT https://beltran12138.github.io/writing/interface-blind-spots https://beltran12138.github.io/writing/interface-blind-spots Freeze CLIP's vision encoder, change nothing about its information, and retrain only a linear map on the text side: binding accuracy jumps from 0.58 to 0.95. The blind spot was never in the encoder. It was in the interface — and no amount of scale touches it. Mon, 13 Jul 2026 00:00:00 GMT https://beltran12138.github.io/writing/understanding-and-fitting https://beltran12138.github.io/writing/understanding-and-fitting Cross-entropy is the only judge in LLM training, and it cannot tell a model that understands from one that merely fits — both produce the same token and the same gradient. Which means understanding is never selected for. It only ever sneaks in disguised as compression. Wed, 08 Jul 2026 00:00:00 GMT https://beltran12138.github.io/writing/the-world-ai-can-compute https://beltran12138.github.io/writing/the-world-ai-can-compute AlphaFold won a Nobel Prize without leaving behind a single readable equation. Three hard constraints suggest that for a large class of high-dimensional systems, compressing the law back into symbols a human can hold is not hard — it is unavailable. Mon, 15 Jun 2026 00:00:00 GMT https://beltran12138.github.io/writing/native-tongue-is-a-manifold https://beltran12138.github.io/writing/native-tongue-is-a-manifold "Which human language suits AI best" is the wrong question. Human language is lossy compression built for a serial human channel the model does not have. Force a reasoning model into a single language and its maths accuracy drops 5.6 points. Wed, 10 Jun 2026 00:00:00 GMT https://beltran12138.github.io/writing/claude-dreams-the-boring-half https://beltran12138.github.io/writing/claude-dreams-the-boring-half Human sleep runs two systems: slow-wave consolidates and deletes contradictions, REM recombines and preserves them. Anthropic built the first half and called it auto-dream. The half they left out already has an engineering name — hallucination — and we spend everything we have suppressing it. Mon, 08 Jun 2026 00:00:00 GMT https://beltran12138.github.io/writing/tokens-are-ceding-the-middle https://beltran12138.github.io/writing/tokens-are-ceding-the-middle A structural audit of discrete tokens against continuous embeddings using three independent mathematical tools — information theory, information geometry, and representation theory. The verdicts disagree, and where they disagree tells you the shape of the next architecture. Wed, 27 May 2026 00:00:00 GMT https://beltran12138.github.io/writing/plato-found-his-wax-tablet https://beltran12138.github.io/writing/plato-found-his-wax-tablet Anthropic cut Claude's blackmail rate to zero not by showing it rules but by feeding it millions of tokens of synthetic fiction about a psychologically healthy AI. That is not alignment engineering. It is fiction authorship — and it is the thing Plato designed 2,400 years ago and could never build. Tue, 19 May 2026 00:00:00 GMT https://beltran12138.github.io/writing/hierarchy-moved-house https://beltran12138.github.io/writing/hierarchy-moved-house Block says AI will replace its management layer. Prussia ran that experiment in 1808, won three wars with it, and watched it fail catastrophically in 1914 — then had the surviving staff officers write the history that exonerated the staff system. Sovereignty does not vanish. It relocates. Mon, 18 May 2026 00:00:00 GMT https://beltran12138.github.io/writing/taste-a-two-hundred-year-old-consolation https://beltran12138.github.io/writing/taste-a-two-hundred-year-old-consolation Silicon Valley has converged on taste as the skill AI cannot replace. Handloom weavers said it about hand-feel, Arts and Crafts said it about the maker's judgment, typesetters said it about the eye. Every time, the narrow technical claim was true. Every time, the top tier fed 1–2% of the trade. Fri, 15 May 2026 00:00:00 GMT https://beltran12138.github.io/writing/the-door-nobody-knew-was-open https://beltran12138.github.io/writing/the-door-nobody-knew-was-open Two uncoordinated engineering decisions accidentally gave GPT-4 a shortcut for reading meaning out of Chinese character shapes. Fixing the tokenizer closed it. The same mechanism is retiring SWIFT MT — and it is how infrastructure quietly deletes capabilities nobody knew they had. Thu, 14 May 2026 00:00:00 GMT