The Computomatix Times

I've 2x’d my productivity as a VP of Product @mercury by creating a "Second Brain" using 5 years of work history, 15k docs with 3.5 million words, and every tool in my stack. It runs locally, is a core part of my every use of LLM, and gets better everyday.
Today, I want to share the stack, the workflow, and the prompt to build it:
Background
I am a VP of Product for @mercury, which is a long way of saying I'm in a lot of meetings, consuming a lot of content across different tools (linear, slack, notion, data analyses), and trying to make sure I actually get stuff done. Working at a company for 5 years and being an information addict, I am essentially a walking encyclopedia for Mercury post 2021-today -- but I've recently found that my scope + workload means I can't keep every plate spinning.
One day, I was scrolling X and came across a series of posts that caught my attention, starting with @tobi's QMD. QMD is a local vector search, and then a few other posts started to show up that connected a few dots for me:
Claude Code launched hooks (per-event prompt injections)
GasTown / OpenClaw launched with the power of orchestrators writing memory + delegating to sub-agents (among many other patterns)
MCPs/CLIs hit a critical mass, and enough of my core tools were available without having to ask admins to give me API keys
@tylercowen did an interview and talked extensively about "writing for AI" in a way that struck a chord - how much output of work already exists that I'm not using?
I decided that it was time to build
Prep work (~1-2 hours end to end)
To start, I needed a library of all the content I could know about... so I downloaded every document I've ever created for my job at Mercury + any relevant product strategy, analysis, retro, reflection on execution, etc. This netted out to over 15k documents and 3.5 million words. Maybe I've read them all, but I've forgotten most. These became a folder that I just called "raw data", and I ran QMD to index this on my computer.
To see if this worked, I used Claude Code to ask about random memories and surprising insights from this knowledge base - the amount of delight/surprise I experienced in seeing how much more capable vector search was than text-based search gave me the confidence to keep going. I asked one questions about books that it would think I like, and it was spooky how good of recommendations it gave me. I think this is my best advice in this journey: test every step of the way! Easy to get caught in hill climbing a local maxima
Train my brain and connect it to my tools (~2 hours)
With all the raw data, I needed to help it make sense of me + what my goals are + the tools I used, so pursued three paths:
Explain myself - to be able to create a second brain, it needed to know what mine was doing. I wrote up a me.md explaining who I am (work + life), gave it my goals + performance reviews for the last 5 years + set of personal priorities. The most humbling part was the system pointing out that I've been making the same strategic mistake for years, according to my own performance reviews, and was making it that week as I was setting up the system
"Distill" the data - I spun up an agent team to use the me.md + the knowledge base to create a set of docs between me <> raw knowledge base. This idea largely came from the idea that LLMs regularly distill down smaller models to take tasks, and I had no idea if it would help me in this, but Agent Teams had just launched and so I had a swarm of them find the main "themes" we've worked on from the knowledge, give sourced histories of this, and summarize key lessons. These created a context.md folder
Tools - I use a few tools (Google Docs, Linear, Notion, Metabase) , and luckily most have connectors on Claude Code or these companies are actively launching MCPs/CLIs. A few didn't, but I spun up specific skills that crafted direct API calls to be able to complete tasks like "run a query for XYZ".
Claude had access to all the information about me + the tools I used + had a massive library of all my work, but did it really know anything? Does anyone?
Wire it up (<1 hour)
At this point, I had so many words + documents that it was time to actually find use or abandon ship. But I didn't want to have to go search this every time and that's when "hooks" caught my attention.
Hooks from Claude Code let you insert content into your prompt without needing to ask (or when a session starts, after a tool use, or when a session stops). Using the UserPromptSubmit hook, I enabled my Claude Code to use qmd to find names + topics + specific documents related to my prompt.
This is a nerd-out moment, but when searching for files in Finder, it is mostly a name + raw text search.... but QMD can help bring context into searches. My system is tuned to figure out a query, then returns results using one of two techniques:
vsearch (semantic/vector) — understands meaning of my question. "How's the funnel performing?" finds … continue on X ↗

Jack Dorsey is a man of contradictions.
He is the only founder to have two companies - Twitter and Block - join the S&P 500. This is an unbelievable accomplishment, surely one of the most impressive in business history.

This article was originally published on my blog. I'll occasionally syndicate them on Twitter but subscribe there if you want all the articles in real time.
educatedguesser.substack.com/welcome


Twitter is a real-time broadcast from your pocket to the world. It is the global nervous system for news, politics and culture. It remains that way today despite new ownership (a true testament to the power of the idea and the network).
Square took the $1,000 payment terminal and compressed it into a $10 piece of plastic that revolutionized small business commerce.
He was CEO of both companies simultaneously. During this period Twitter was famously described by Mark Zuckerberg as a “clown car” while Block let Toast, Stripe, and Shopify steal its lunch right out from under its nose. Both companies became bloated, sprawling fiefdoms and were eventually gutted (Twitter, famously by Elon Musk, and Block/Square/XYZ by his own hand). The divided focus did not work.
It’s become fashionable over the last few years to use Jack’s track record of executional missteps to dismiss him, and his ideas, entirely.
And to be fair, it hasn’t been the prettiest few years:
He bought Afterpay at an announced price of $29B (at least he used stock for the acquisition). Block’s market cap four years later? $38B.
He bought Tidal. Tidal! I think there was a reason besides being friends with Jay-Z but I can’t remember it.
Elon cut 80% of Twitter and the team ships faster today than they ever did during the Dorsey Era.
Oh, and we can’t forget the time he turned himself into a literal blockhead.

People struggle to hold these two Jacks in their heads at the same time. And after the last few years, they focus on the execution missteps and dismiss the innovator who is able to see the future, pull it forward, and put it in your pocket before people even realize the world has changed.
What they don’t realize is that these two sides of Jack Dorsey are two sides of the same trait.
The Jack who can’t sit still long enough to rigorously run a mature organization is the same Jack looking out five years, realizing the world will be radically different, and taking the knife to his own company. The Jack that lets his companies get way too big is the same Jack who can recognize the structure is now a noose in the AI era, and cut 40% in one go while his peers cut 10% each year and call it performance management.
Introducing: Dorsey Mode
Given Jack’s track record, I listened with real interest to his recent appearance on @bhalligan Long Strange Trip, where he and Roelof Botha deconstructed what Halligan is now cheekily calling Dorsey Mode, Jack’s radical new approach to management in the AI era.
youtube.com/watch?v=YTVSwOY19Qs
I’ll be honest. When Block announced the 40% layoff, I dismissed it. You can read what I said on Twitter right after the news dropped. I indexed way too hard on “unfocused” Jack without considering “visionary Jack.”
I said it had nothing to do with AI. I was wrong.
After listening to the full conversation, I’ve updated my position. Jack is pulling forward the future again and rebuilding his company for where AI is going to be.
He’s done this a few times now, and people always laugh at him. But more often than not, he’s right. Hell, the fact that his ideas keep working despite his execution probably means the ideas are twice as powerful as we give them credit for.
So here’s my updated read on Dorsey Mode, the four parts of his thesis that I think actually matter, and why I think Jack is early and right. Again.
The Four Big Ideas Behind Dorsey Mode
1. Cut 40% now
Brian Halligan: You laid off 40 percent of your employees. You know, Ruth Porat’s got this good line—if you’re gonna eat a shit sandwich, don’t nibble.
Jack Dorsey: We’d been making changes on the edge, like going from a GM structure to a functional structure to reduce—like, putting a cap on our layers to four—me plus four—and all these small things. But if we were to really reboot and rebuild the company, would we end up where we look today? And the answer was uniformly no.And I think generally I wanted to make sure that we—if we knew that this was what our company was going to be in the future, I didn’t want to have to do it with our backs against the wall. We’re a public company, and there’s various challenges there. And other companies will probably get to this realization at some point. I don’t want to react to that.I want to be ahead of it, because then we can do it with a lot more integrity. We can do it with a lot more generosity for the people that we’re asking to leave, and even for the people that we’re asking to stay. And we’re not just reacting into something mediocre. We’re acting towards excellence. And that’s just the tone t… continue on X ↗

I've been building iOS and macOS apps with Codex and Claude Code since last year. One thing I've learned: agents need simple systems that produce reliable results.
Without the right skills, agents write deprecated code, create compiler errors, and waste your time debugging their mistakes. These 5 skill packs fix that. Each one comes from a developer who's been shipping real apps with agents.
1. Paul Hudson: The Swift Foundation

If you've learned Swift online, you've probably read Paul Hudson's work (@twostraws). His agent skills come from a decade of teaching Swift through Hacking with Swift — and his SwiftUI rules help correct common mistakes with agents.
I met Paul at try! Swift in NYC and we spent some time together in the SwiftUI labs at WWDC 2019. He's a prolific writer and digs in deep so you can understand what works.

These skills will jumpstart your agents:
SwiftUI Pro: modern SwiftUI APIs, view composition, state management
Swift Concurrency Pro: async/await, actors, Sendable, Swift 6 migration
Swift Testing Pro: Test macros, parameterized tests, XCTest migration
SwiftData Pro: models, queries, migrations, CloudKit sync
He also maintains the canonical directory of every community Swift skill.
2. Antoine van der Lee: SwiftLee

Antoine van der Lee (@twannl) runs SwiftLee — one of the most widely read Swift blogs. He has 5 skill repos, each with detailed reference docs. His Xcode Build Optimization skill stands out — 6 sub-skills for build settings and compilation times to make your code build faster.
SwiftUI Expert: state management, view composition, performance, Liquid Glass
Swift Concurrency: actors, Sendable, data race safety, Swift 6 migration
Swift Testing Expert: modern testing patterns, XCTest migration, parameterized tests
Core Data Expert: stack setup, fetch requests, background contexts, migrations
Xcode Build Optimization: 6 sub-skills for build settings, compilation times, project config
The Xcode Build Optimization skill is a powerful tool that can make both you and your agents more productive. And if you want a better workflow with the iOS Simulator, check out his developer app: @rocketsim_app
3. Thomas Ricouard: Codex Expert

Thomas Ricouard (@Dimillian) built Codex Monitor — the open source macOS app for managing multiple Codex agents — and then joined OpenAI's Developer Experience team.
His skills now ship as the official Codex Build iOS App and Build Mac App plugins. These automatically update with each new release of the Codex app or Codex CLI.
Build iOS Apps plugin
SwiftUI Liquid Glass: iOS 26+ Liquid Glass APIs, modifier ordering, fallbacks
SwiftUI UI Patterns: navigation, sheets, app wiring, reusable components
SwiftUI Performance Audit: invalidation storms, identity churn, layout thrash
SwiftUI View Refactor: smaller subviews, MV-style data flow, Observation
iOS Debugger Agent: simulator build/run/debug with XcodeBuildMCP
iOS App Intents: Siri, Shortcuts, widget integration
Build macOS Apps plugin adds AppKit interop, packaging/notarization, signing/entitlements, window management, and SwiftPM workflows.
You can dig into the source for both of these plugins at the Official OpenAI Plugins repo,
4. Krzysztof Zabłocki: Advanced Swift + Tooling

Krzysztof Zabłocki (@merowing_) created Sourcery — the Swift metaprogramming tool used by 40,000+ apps including Airbnb and The New York Times. His open source tools power over 80,000 apps total.
His approach to agent skills is completely different. Instead of individual skill files, he built a rules-based system over 3 years of daily LLM use — 12 domain-specific rule files with a smart loader that uses LLM self-reflection to decide which rules to apply based on context. The system is tool-agnostic: same rules work in Cursor, Claude Code, and Codex.
What I like about his approach: he has a progressive documentation reading CLI tool to offload search and make it agent-friendly. He also created Inject for hot Swift reloading — useful for fast prototyping when you want to see UI changes without rebuilding.
Read his guide: Stop Getting Average Code from Your LLM
You can grab two agent friendly files that follow his coding philosophy:
general.md: primary directive and coding standards
rule-loading.md: smart loader that selects rules by context
The full 12-file set covering dependency injection, SwiftUI architecture, ViewModel coordination, and Swift Testing is part of his Swifty Stack course.
5. AppCreator: Agent-Friendly Build Tools
None of the skills above matter if your agent can't build and test your code.
Xcode build output is verbose. Test output is worse. We're juggling XCTest and Swift Testing with two different build systems. Agents choke on this.
I built AppCreator to solve that problem. It scaffolds Xcode projects with agent-friendly defaults. I created it to quickly prototype new app ideas that used the same workflow as my existing projects.
Adopt the skill for your existing apps and it will help your agent make a build … continue on X ↗