Compounding

I think a lot about compounding.  Not in the sense of compounding interest, tho that is a great and powerful thing.  But the compounding of knowledge.

One of my favorite books is Jeff Tweedy’s “How to Write One Song”.  He describes how he just keeps plugging away, building an ever-increasing stack of lyrics and melody ideas day after day.  Writing music, writing a book, building products, educating yourself — they are all about continued, diligent effort, adding one brick at a time.  

Jeff uses a simple system to manage his lyric and melody ideas.  And both “simple” and “system” are key for any work — you need a repeatable system that you can use every day to build and extend your knowledge, and it has to be simple, or you won’t use it, or you will spend too much time fussing with it.  

I dabbled in tools like Obsidian and Notion for note keeping, and I found that they actually worked against the compounding of knowledge — they were not simple enough, they were too fussy about how you structured your notes, and it was easy to spend way too much time dorking around with the tools, and not enough time on working and learning.

Andrej Karpathy has proposed what seems in retrospect an obvious idea — let an llm help manage the structuring and connecting of your knowledge:

The tedious part of maintaining a knowledge base is not the reading or the thinking — it's the bookkeeping. Updating cross-references, keeping summaries current, noting when new data contradicts old claims, maintaining consistency across dozens of pages. Humans abandon wikis because the maintenance burden grows faster than the value. LLMs don't get bored, don't forget to update a cross-reference, and can touch 15 files in one pass. The wiki stays maintained because the cost of maintenance is near zero.

I kind of love this idea.  It provides real muscle to the idea of compounding your personal knowledge.

Developers

I always thought Microsoft’s 2018 acquisition of GitHub was a smart move.  GitHub was at the center of every developer’s workflow and every startup's workflow.  Microsoft had always had developers as its core customer focus, and this acquisition was a smart move by Satya to re-establish Microsoft as the premier developer technology company.   Unlike in past acquisitions, Microsoft handled GitHub gently; GitHub continued to be a great service for all developers, not just Microsoft-centric ones.

But the developer market has moved on.  Despite early investment, Microsoft has not kept up with the changing developer workflow.  The Pragmatic Engineer has a very good article about Microsoft’s slippage, with some damning observations:

GitHub and Copilot are entangled with Microsoft’s internal politics. GitHub’s Copilot in 2021 was the first massively successful “AI product.” Microsoft took the “Copilot” brand and used it across all of their product lines, creating low-quality AI integrations. Simultaneously, internal Microsoft orgs like Azure and Microsoft AI were trying to get their hands on GitHub, which is one of the most positive developer brands at Microsoft.

GitHub has no leader, seemingly by design. GitHub’s last CEO was Thomas Dohmke, who stepped down voluntarily, and Microsoft never backfilled the CEO role; instead carrying out a reorg to make GitHub part of Microsoft’s AI group and stripping its independence. It seems the “Microsoft AI” side won that battle.

GitHub has no focus, and is stuck chasing Copilot as a revenue source. GitHub has no CEO and is caught up in internal politics, so, what can GitHub teams do? The safest bet is to increase revenue and the best way to do that is by investing more into GitHub Copilot, and ignoring long-term issues like reliability.

and

GitHub Copilot went from the most-used AI agent in 2021, to be overtaken by Claude Code, and is soon to be overtaken by Cursor.

As a platform, GitHub has no vision for how to evolve to support AI agents. Sure, GitHub has an MCP server, but it has no “AI-native git platform” that can handle the massive load AI agents generate.

This should be a five-alarm fire inside of Microsoft.  As with Copilot user adoption, Microsoft has clearly lost the thread.  

Two years ago, I couldn’t have imagined developing without an IDE like VS Code.  Now VS Code seems increasingly stodgy and in the way.

Compounding at the National Level

How will a country spend its time and attention?   Will it build the infrastructure and systems that result in compounding growth?

As the US is distracted by wars and internal politics, China is just laughing at us, as they spend on roads and trains and airports and foreign aid and basic research and electrification and more.  And I can’t complain about China — they are doing great things for their country and for humanity.  An explosion of Chinese-funded R&D will eventually help us all.

For the US, deciding how to work with China is our number one foreign policy issue for probably the next century.  And so what are we doing?   We are eating through our federal budget, blowing things up in Iran.  The Administration now wants a 40% increase in the defense budget, and we are still in the early days and years of this conflict.  

I’m finally reading “All the Shah’s Men”, an excellent telling of the US’s complete bumbling in Iran from the 1950s on.  And now we are compounding the errors, creating an entirely new generation of martyrs and hatred for the US.  We will be paying for this for decades, while we do little or nothing to improve our competitiveness with China.