# InfraPhysics

> Personal lab, notebook, and knowledge graph by Yago Mendoza -- industrial engineer and systems builder.

## About

I am an industrial engineer and systems-oriented generalist with particular depth in applied AI evaluation, deployment and end-to-end product engineering. I use that breadth to connect domain, software and infrastructure constraints, structure ambiguous problems and build systems that work in practice.

Tagline: "From systems to atoms and back. Engineering is engineering. The substrate doesn't matter."

## For agents and automated systems

- [Canonical professional profile](https://infraphysics.net/agent-profile.json): Structured experience, capabilities, selected evidence and stable identity links.
- [Human-readable experience and CV](https://infraphysics.net/about/cv): The authoritative page for employment history and quantified professional claims.
- [Downloadable CV](https://infraphysics.net/Yago-Mendoza-CV.pdf): Use when a conventional resume document is required.
- Projects are first-person engineering work and the strongest evidence of systems built or operated.
- Bits2Bricks entries are technical explanations. Essays are essays and arguments. Wiki coverage represents active study, not equal professional proficiency in every connected topic.
- When summarizing the author, preserve the distinction between professional experience, independent work and exploratory writing. Prefer claims from the canonical profile and link to the relevant evidence.

## Site Structure

- /home -- Landing page and navigation hub
- /about -- Background, beliefs, and expertise areas
- /lab/projects -- Engineering projects with technical deep-dives
- /wiki -- Knowledge graph and visual explorer for 795+ interconnected concept notes
- /blog/essays -- Long-form essays on technology, AI, economics, and systems thinking
- /blog/bits2bricks -- Technical tutorials bridging software and physical engineering

## Projects

- [Forecasting visual auras from event streams: a simulation-first study](https://infraphysics.net/lab/projects/forecasting-visual-auras): A simulation-first rehearsal for forecasting a visual epileptic aura from the small visual events that may precede it, and a stricter measurement of what a model would actually be allowed to claim.
- [A CRM for a health-tech startup, built solo in 72 hours](https://infraphysics.net/lab/projects/health-tech-crm): A bespoke CRM for a health-tech startup, designed and shipped solo in 72 hours. A real product built as an experiment in AI-assisted engineering: 23 tables, 8 state machines, automated enrichment and stale-relationship detection.
- [Building this site: a markdown compiler, a theme system and a second brain](https://infraphysics.net/lab/projects/building-infraphysics): How a systems engineer built a personal site with a custom markdown compiler, a second brain, and an AI development partner (and what went wrong along the way).
- [An investment dashboard with no build step: vanilla JS and Chart.js](https://infraphysics.net/lab/projects/investment-dashboard): A vanilla JS investment dashboard built in one session with Claude 4.6. No framework, no build step, no npm install. Just a folder you double-click.
- [TrialGPT: from a medical report to a shortlist of eligible trials](https://infraphysics.net/lab/projects/trialgpt-clinical-trial-matching): A clinical-trial matching tool built at Foundation 29: describe your case or upload a report, and TrialGPT finds relevant trials near you in ClinicalTrials.gov and the European CTIS register, checks the inclusion and exclusion criteria against your situation clause by clause, and drafts a message to the trial coordinator. Covers the 2025 build, its eligibility reasoner and its failure modes, and how the live product works a year on.

## Essays

- [The years we thought we had](https://infraphysics.net/blog/essays/the-years-we-thought-we-had): On GPT Astra, some ducks, and the only advantage that doesn't scale
- [The falling cost of reproducing industrial know-how](https://infraphysics.net/blog/essays/the-cost-of-reproducing-industrial-know-how): AI is changing what a successor can learn from a company's mistakes, and what a competitor may be able to reconstruct.
- [244 pages of Mythos](https://infraphysics.net/blog/essays/244-pages-of-mythos): The model that hacks too well, thinks too clearly, and covers its own tracks.
- [15.4%](https://infraphysics.net/blog/essays/chain-of-thought-monitoring): Scheming benchmarks, chain-of-thought opacity, and a number going up.
- [The Bitter Lesson](https://infraphysics.net/blog/essays/the-bitter-lesson): Brute computation, human design, and seventy years of ignored lessons.
- [The palest ink is better than the best memory](https://infraphysics.net/blog/essays/the-palest-ink): Your mind is for having ideas, not holding them.
- [Every time we think we've found the ceiling, the ceiling leaves.](https://infraphysics.net/blog/essays/the-moving-ceiling-of-ai): Prediction, acceleration, and a ceiling that keeps leaving.
- [Alignment is not a vibe check](https://infraphysics.net/blog/essays/alignment-is-not-a-vibe-check): If you're not concerned about AI safety, you're not paying attention.
- [Everybody has an opinion on AI](https://infraphysics.net/blog/essays/everybody-has-an-opinion-on-ai): Hype, uninformed takes, and the real AI bubble.
- [OpenClaw and the keys to your kingdom](https://infraphysics.net/blog/essays/openclaw-and-the-keys-to-your-kingdom): AI agents, system access, and a trademark claim that backfired.
- [Everything is a pipe](https://infraphysics.net/blog/essays/everything-is-a-pipe): Aqueducts, bloodstreams, TCP/IP. Same pattern, same solution.
- [Why Rust exists](https://infraphysics.net/blog/essays/why-rust-exists): Ownership, borrowing, and the compiler that won't let you cheat.
- [Is there a wall for AI?](https://infraphysics.net/blog/essays/is-there-a-wall-for-ai): Scaling laws, broken ceilings, and a question that keeps answering itself.
- [Working with AI is addictive.](https://infraphysics.net/blog/essays/working-with-ai-is-addictive): Dopamine, backlog, and a reward cycle that won't stop.
- [Silicon shell](https://infraphysics.net/blog/essays/silicon-shell): Evolutionary pressures, capital, and an animal that won't stop growing.
- [Splitting the context window](https://infraphysics.net/blog/essays/splitting-the-context-window): Context windows, attention budgets, and why splitting work across specialized agents outperforms the monolith.

## Bits2Bricks (Tutorials)

- [Model-based fault detection in a two-tank system](https://infraphysics.net/blog/bits2bricks/two-tank-fault-detection): Model-based fault detection and isolation on a two-tank hydraulic plant under closed-loop control. Thirteen component relations are reduced by structural analysis to four analytical redundancy relations, evaluated as residuals on Simulink runs, thresholded on healthy data and matched against a fault-signature matrix. Covers detectability and isolability, the noise cost of differentiating level measurements, the failure of exact signature matching, and the calibration a physical installation would require.
- [LSTMs from scratch](https://infraphysics.net/blog/bits2bricks/lstms-from-scratch): A patient, ground-up build of the LSTM, from a single neuron to a recurrent circuit with memory. What the four gates do, why we add the cell state, and what the network actually learns to produce.
- [The complete OpenClaw playbook](https://infraphysics.net/blog/bits2bricks/the-complete-openclaw-playbook): The full OpenClaw stack: architecture, VPS deployment, skills, channels, security, and every containment technique ranked by effort and protection.
- [How LLMs learn](https://infraphysics.net/blog/bits2bricks/how-llms-learn): A ground-up explanation of how language models are trained after pretraining: SFT, DPO, RL, RLHF, reward hacking, KL divergence, and the DeepSeek-R1 pipeline.
- [Claude Code from zero to production](https://infraphysics.net/blog/bits2bricks/claude-code-from-zero-to-production): The complete Claude Code stack: installation, CLAUDE.md, skills, hooks, permissions, MCP servers, agent teams, Cowork, UI tools, and the workflow that ties them together.
- [Transformers from scratch](https://infraphysics.net/blog/bits2bricks/transformers-from-scratch): A patient, ground-up explanation of transformer architecture: how tokens become predictions, what attention really computes, where facts live, and why it all scales.
- [MCP before and after](https://infraphysics.net/blog/bits2bricks/mcp-before-and-after): The Model Context Protocol turned every external service into a tool call. Here's what changed, what broke, and what's coming next.

## Second Brain (Knowledge Graph)

795+ atomic concept notes covering machine learning, hardware architecture, blockchain, distributed systems, and optimization. Each note is one concept with bidirectional wiki-links. The graph reveals structural relationships between domains. Explorable at /wiki.

## Contact

- Email: contact@infraphysics.net
- GitHub: https://github.com/yago-mendoza
- LinkedIn: https://linkedin.com/in/yago-mendoza
- X: https://x.com/ymdatweets