Shawn
_@shawnyeager.com
npub1clk6...pup9
From product to pipeline—GTM advisory for frontier tech
Bitcoin · AI · Distributed Systems · Open Source
Podcast: trustrevolution.co
Substack: sideband.pub

Big fan of Immich. I replaced iCloud Photos with it last year. The founder just mentioned on their discord that they're hiring a mobile dev. 🚀


Mobile Developer - Immich
Apply for Mobile Developer - Immich at FUTO.
Interesting thing I noticed with Claude Code: I use the compound-engineering plugin (https://github.com/EveryInc/compound-engineering-plugin) which documents solutions in `docs/solutions/` per repo.
During a planning session in a new repo, the agent couldn't find local learnings, so it scanned `~/Work/` for `docs/solutions/` across all my other repos and pulled relevant patterns from two different projects.
Nobody told it to do that—the prompt just said “check any other relevant learning directories.”
Emergent cross-repo knowledge sharing from a vague instruction.
Let's stir up the hornets.
Agent autonomy requires native machine-to-machine payments in sats. No humans in the loop or fiat rails.
Lightning + L402 is the obvious technical fit for that. The lightning-agent-tools release shows agents can already pay and get paid autonomously. (Note: While I'm bullish on Cashu, no mid to large enterprise is going to touch it any time soon.)
Today, fiat rails are winning and will keep winning for the foreseeable future. Volatility, treasury accounting, node/liquidity ops, and regulatory optics make native Lightning too heavy, while Stripe and stablecoin wrappers are simple, stable, and already approved.
Long term, Lightning is the only rail that makes sense for a sovereign agent economy. Short to medium term, ease beats independence for most orgs.
I want to be proven wrong. Where are companies or teams running production agent-to-agent payments natively on Lightning today at any real scale?
College-age me is stoked.


AI isn’t killing software—it’s forcing a brutal bifurcation. Half the industry (simple workflow tools, basic automation) gets unbundled and commoditized. The other half (orchestration platforms, vertical depth, security/infra) gets supercharged if they adapt fast.
Go-to-market is now the real moat. Building is table stakes; getting people to pay reliably in an AI world is the new hard problem.
Trusted > Followed
Replacement theory.


Threads
Every major technology frontier is a race between the people building to empower and the people building to extract. The extractors are usually faster.
The future belongs to people who touch it with their own hands, not people who read about it.
The people building the most important technology and the people who know how to sell it rarely overlap. That gap is where the most indispensable work happens.
You can see the future right now—if you're willing to build, not just consume.
During dinner last night, my friend, who’s franchising his business, explained the mafia tactics of brokers and sales organizations.
So, I came home, stayed up a little too late, and started building with Claude Code. Less than 24 hours later, I built tender for franchises. I did it, as I have with many recent projects, to push the tools and prove to myself what's possible.
Turns out, more and more each day.



