I am a bay area-based software engineer and researcher in previous lives I have [programmed gold mines](http://www.hanlonengineering.com/) , [written high-scale web crawlers](), written [a distributed database](https://www.citusdata.com/) which pretends to be postgres, scaled [a COVID testing lab](https://curative.com) up to 10% of all US testing, designed [network protocols for adversarial environments](https://ethereum.org/en/), and shipped [what was the cheapest way to send crypto, by an order of magnitude](https://world.org/) along the way I've spent countless hours with profiling tools to build "mechanical sympathy'", an understanding of what some system is really doing and how to make best use of it. until you have specifically looked and made changes your program is probably doing something silly and inefficient. last fall I spent some quality time with nsight to write [a little latent space explorer](https://x.com/bmc_/status/1846361357266178245) which ran SDXL-Turbo inference faster than any provider I could find recently I am thinking about a different kind of system, watching the field of mechanistic interpretability with fascination and trying to understand what it would mean to "profile" gemma, to understand each of the circuits and how they are regulated. is each of those billions of parameters truly necessary, or is gemma also doing something silly and inefficient? further afield, maybe with the right kind of tooling we can discover that sub-circuits which compute "what task am I performing" exist, prevent them from ever attending to user-supplied tokens, and thereby eliminate prompt injection? I also enjoy zooming out. some fascinations as of time of writing: the philosophy and methods of science, coordination mechanisms for large groups, [the possibility of directly imaging extra-solar planets](https://www.nasa.gov/general/direct-multipixel-imaging-and-spectroscopy-of-an-exoplanet-with-a-solar-gravitational-lens-mission/), the future study of [[Are there any linguistic universals|xenolinguistics]] if you made it this far and think you would enjoy talking to me, I would probably enjoy talking with you. you are reading this on briancloutier.com, your first guess as to my email address is correct in case any of these provide further inspiration for reaching out here are some recent favorite books: - [Inventing Temperature](https://global.oup.com/academic/product/inventing-temperature-9780195337389) | it is surprisingly difficult to establish that water freezes at 0° and boils at 100° and use this fact to calibrate thermometers and then use thermometers to invent a theory of heat. the biggest difficulty is that water absolutely does not freeze at 0° nor boil at 100° - [Designing an Internet](https://direct.mit.edu/books/book/4216/Designing-an-Internet) | DDOS attacks are only possible because of specific decisions we made when designing the internet, we could have designed it differently. [active networking](https://en.wikipedia.org/wiki/Active_networking) is discussed here and is particularly exciting (what if every packet was a program?) but it will never happen, for exactly the same reasons as why I am excited by it - [The Vital Question](https://nick-lane.net/books/the-vital-question-why-is-life-the-way-it-is/) | our mitochondria might explain why there are two sexes, and also everything else about us (obviously they don't, but I love grand books which try to take one theory and stretch it to explain everything. see also: [Scale](https://www.penguinrandomhouse.com/books/314049/scale-by-geoffrey-west/)) - [Governing the Commons](https://www.cambridge.org/core/books/governing-the-commons/A8BB63BC4A1433A50A3FB92EDBBB97D5) | under certain conditions it is possible to manage communal resources using neither government control nor private ownership; "regulation vs laissez faire" is a false dichotomy - [Who gets What and Why](https://www.harperacademic.com/book/9780544705289/who-gets-what-and-why/) | governments are uniquely well suited to provision public goods by first designing a market and regulating it such that it leads to efficient outcomes then letting that market do that rest. "regulation vs markets" is also a false dichotomy