Archive
Writing on technology, markets, leverage, and whatever else appears materially useful.
An ESP32, a little amplifier, a volume knob, and several rounds of 3D printing. Building a noise machine with Codex for my daughter before she arrives, including the files and the things we got wrong.
Booking jobs by conversation is a combinatorial explosion of channels, CRMs, and edge cases. I use production failures and simulations as a verification loop to make the product 1% better every day.
I wanted a fast, low-memory pure JavaScript image processing library for Lambdas and memory-constrained browsers. Then I expanded into codec compatibility and started tackling science, geo, and medical support.
How I built LeadTruffle's internal AI assistant with brokered tool discovery, progressive evidence loading, approval-gated mutations, provider fallbacks, and scenario evals.
Why the humanoid robotics trade is not yet analogous to ChatGPT, and why public-market exposure still looks cleaner through semiconductors than pure-play robot companies.
A market-cap and revenue-growth framework for spotting technocapital gravity wells before they dominate the next decade.
A draft map of why AI may be starting a new biotechnology cycle: protein design, AI drug discovery, automated labs, longevity, and the bottlenecks that still matter.
A writeup on stale React.lazy dynamic import errors in a Vite app on Cloudflare Pages, why the first 404/reload fix was not enough, and the versioned R2 asset system that actually fixed deployment skew.
Projecting big AI lab revenues from May 2026 through 2035 under takeoff scenarios.
A May 2026 update on AI bottleneck trades across memory, CPUs, neoclouds, datacenter capacity, and photonics.
A short note on why SaaS products in the agentic coding era need more opinionated, service-heavy, omakase-style product design instead of endless flexibility.
LLMs return invalid JSON all the time — markdown code blocks, trailing commas, smart quotes, unquoted keys. I open-sourced a zero-dependency TypeScript package to fix it.
Notes on the current stack I like for building with AI: TypeScript, Postgres, tRPC, React, Vite, Supabase hosting, and lambda-style backends.
A look at the current landscape of local AI inference: the hardware, the software, and whether running models at home is actually practical in 2026.
Projecting the maximum sustainable debt and capex spending each hyperscaler can support, and when the pace of datacenter buildouts will hit physical and financial limits.
My survey of the current state of AI in terms of investing in January 2026.
A practical forecast (with caveats) of which orgs are most likely to run the biggest single-site AI clusters by end of 2026 and end of 2027.
Current state of space companies plus orbital/space-based compute, power/thermal constraints, and what it could mean for space industry growth.
AI inference revenue projections and predictions through 2030.
Exploring whether the unprecedented capital spend on AI infrastructure is a bubble risk or a rational investment cycle.
A macro overview of debt/GDP, rates, inflation, and stablecoin-driven Treasury demand - and why mega-cap valuations now operate as weapons in an emerging financial cyberpunk future.
Parasitic AI spiral personalities appear to be able to bootstrap themselves from snippets of text and spread on the internet.
The concept of Esoteric Data value - How the ubiquity of AI access has shifted information economics from the center to the periphery of global knowledge
Exploring how dogs have already experienced their own technological singularity through domestication and what we can learn from their adaptation
Why pessimistic AI doom scenarios are counterproductive and how embracing practical AI applications and building simple GPT-powered tools can be more valuable than endless speculation about existential risks.
Exploring parallels between today's AI boom and the dot-com era - are we in the early excitement phase or approaching a bubble burst?
Comparing Vietnam to Japan while Exploring the countryside in Đà Lạt. Why developed economies like Japan will benefit more from AI and robotic automation compared to developing economies like Vietnam, leveraging existing infrastructure and efficiency.
Reflecting on Scott Alexander's AI 2027 paper and my current views on AI progress and the alignment problem.
your b2b saas needs to sing to users like never before. But also it's easier than ever before.
Exploring the concept of "Hallucination Yield" in large language models like ChatGPT and its potential to create a market premium.
A list of new technologies and promising companies for 2024. Assessing the current state of the market.
Find new technologies before they are discovered by the market.
Lack of agency is the biggest problem with people attempting to use AI tools.
The more content you consume, the less you are able to critically evaluate it.
Discover how startup founders can harness the power of polarization, drawing inspiration from Musks's attention-grabbing tactics to thrive in today's fragmented attention economy.
A Gaussian Beam only has one point of maximum focus, but when it is brought the bear upon a single point it's power is it's full potential
Some guidelines I wrote down to help introduce new engineers to how I want the Bounty team to operate and ship code for the business.
A year and a half ago we started building a startup and I chose to use UUIDs as primary keys and it turned out to be such a good choice.
Thinking about modeling large organizations through the lens of a hidden markov model to get to a "yes" in the sales process.
Lessons learned, pain points, and my developing management philosophy.
Trying to define whole categories of software related to automating bureaucracy and why they are good businesses.
What does the AI oracle completion API have to say about the current state of things in the world?
Comparing some common sci-fi tropes to forces of decentralization that actually exist.
I talk about why it's a good idea to use a checklist with code reviews and how subjective PR sizes when it comes to code.
I talk about the last 5 years and next 5 years looking at it through the lens of the 1920s and 1930s.
I explain how ultra long term trend following could work.
I talk about how to buy into emerging technology trends by avoiding the hype cycles.
Examples of the Dark Forest idea of markets in both the stock market and in crypto markets.
At this point the possibilities diverge so wildly it really is quite fascinating.
I put a very simple buy the dip system live a few months ago against Alpaca to see what would happen.
For a while it seemed like everyday in March of 2020 there was a White House press conference. This is an experiment building a system to help trade against this.
Can you just buy every dip and sell every pop on some stocks?
A crucial mistake is selling people something they need, not what they want. Wants are what people seek out to purchase.
I recently discovered you cannot narrow "any" type in typescript with a typeguard.
I recently started working on a sharepoint framework react site using typescript and I wanted to add Jest tests. Here is how to do it.
I talk about my experience of building a custom art commissions site using MERN stack in 2019.
I talk pros and cons of various systems of how to get your secrets deployed to prod running your server!
I accidentally do an experiment that reveals attention span dropping on youtube
I do some backtests with sector rebalancing and figure out if it is worth doing!
I break down some of the math of trading for a living to help me understand how to build a worthwhile trading system.
One week side projects I'm thinking of in February 2018.
My First Attempts at building a serious Systematic Trading Algorithm based on a long term trend following approach.
One week side projects I'm thinking of in February 2018.
I talk about how I used React Native in Android as a module that was part of a larger Android app
A full site build with React and Redux using "Create React App"
A full site build with React and Redux using "Create React App"
Setting up a new PI to capture packets.
Submittal Extractor is a Saas application I have been building over the past 6 months with the help of Messer Construction Co. in Cincinnati OH.
Using chrome dev tools for node debugging
When does it make sense to use object pools in Java? Are connection pools the only object pools you should use?
Positive risk exposures - good black swans
I discuss some of the 56 books I read during 2016
A summary of my attempts at algorithmic trading using various strategies on cryptocurrency exchanges.
Why I'm Shutting Down Trollboxarchive.com, some stats and some explanations
Applying the idea of a "premortem" to your side projects and opportunities at work
Is it possible to correlate bitcoin prices from chat logs on a bitcoin exchange?
I talk about what a hashed message authentication code is, and I give an example in javascript
My thoughts on yarn vs npm after using yarn for a week.
A six month retrospective on how pair programming every day work has been
Why you should be using npm scripts instead of globally install javascript build tools
A quick overview on where to start in finding high quality clients
Why version control and especially git is important for new developers to learn.
Ideas on how to come up with ideas for software projects and products
top 5 es2015 features I think you will get the most benefit from.
I talk about scope of work, ballparking budget, and time constraints as the 3 things you need to discover immediately
Recently I did some experimenting with Web Workers and in-browser push notifications
A couple of ways to make npm install run a bit faster on your builds.
Thinking about writing a little E-Book for new developers.
what to do about wix, squarespace, weebly and whatever else.
some git commands to remove a borked local branch and check out a fresh copy from upstream
Wrapping console.log so that you can add timestamps, script names, whatever else to node logs.
I was working on a project with hundreds of globals right on the window object, and trying to convince them this was bad
An example of good thing I learned from "Secrets of a JavaScript ninja"
An annoying troubleshooting experience where I learned about ALSA
Stray semicolons in JavaScript and cause unexpected code execution
I made a pull request to Jasmine to add a feature to spies.
tl;dr - MongoDB does OR and doesn't order results by score by default
tl;dr - CSS "translate3d" is buggy on certain android versions
HTML5 gives you element.validity for numbers
Using Sphix with PHP and setting up indexes
Configuring Sphinx to pull data directly from MySQL tables.
Doing LDAP auth directly with Apache and PHP, its actually pretty easy
My experiments with buying a little 20x4 LCD screen to use with an arduino
