In partnership with
Thanks to reader Pietro who asked me to write about skills to develop in order to have leverage in 2025.
If you’re an engineer or early-stage founder, you’re hearing the buzz: AI will eat software. Karparthy and others are saying AI is the new electricity: https://www.youtube.com/watch?v=LCEmiRjPEtQ
I’ve chased the shiny stuff - demos, devtools, Twitter threads with too many rocket emojis. But if you're playing for real leverage - not likes - you need more than hype.
Here’s how we thought about it at The Firm, and how I’m thinking about it now, 2 years post-leap:
Situational awareness → Ruthless prioritization → Repeated, boring execution.
Let's break this down.
🎯 Step 1: Pull the Right Levers
You don’t need to “do AI.” You need to use it where it matters.
Data moats beat model weights
Everyone’s fine-tuning. No one’s asking: what data do we uniquely own? Internal process logs, partner integrations, fine-grained event telemetry - that’s leverage.
Workflows where mistakes cost money
Think: triage queues, L1-L2 escalation, fraud ops, pricing tweaks. If your AI screws up, someone pays. That’s where adoption sticks.
Build legos, not cathedrals
Agents, APIs, SDKs > full-stack monoliths. If someone can plug your thing in and 10x a pain point, they won’t care how fancy your LLM wrapper is.
StartEngine’s $30M Surge — Own a Piece Before June 26
Private markets are having a moment, thanks to companies like StartEngine.
The leading alternative investing platform is helping everyday investors like you access deals once reserved for VCs and insiders, including exposure to private market titans like OpenAI, Databricks, and Perplexity.¹
How’s it going? In Q1 2025, StartEngine pulled off $30M in revenue, its biggest quarter ever (based on unaudited financials).²
But StartEngine isn’t just a middleman. The company earns 20% carried interest on select pre-IPO offerings, unlocking value for shareholders when these deals succeed.³
How can you tap into this diversification play? By investing in StartEngine.
StartEngine has crowdfunded $85M+ to date, and you can join 45K+ shareholders before the company’s current round closes on June 26.
Reg A+ via StartEngine Crowdfunding, Inc. No BD/intermediary involved. Investment is speculative, illiquid & high risk. See OC and Risks on page.
🛠️ Step 2: Engineer for Leverage
You don’t scale with tools. You scale with systems.
Find the 20% of AI use cases that tie to revenue or retention
I once built a model that boosted weekly engagement by 12%. Not because it was smart - but because it saved people clicks. Don’t overthink it.
Make failure visible and boring
Log it. Alert on it. Test it like backend code. Most AI fails like a toddler with a sugar rush - loud, messy, unpredictable.
Automate yourself out of the loop
Retraining, eval, rollout gating - turn it into bash scripts. Your job isn’t to babysit the model. It’s to build the rails it runs on.
🧗 Step 3: Pick a Stack Position with Real Leverage
What you build matters. Where you sit matters more.
Engineer - Operator
You ship and support. You get paged. You choose what not to build. You're not just technical - you’re trusted.
Founder with a wedge
Start with a painkiller. Compound with AI. You don't sell AI - you sell less churn, faster throughput, better conversion. AI is just the sauce.
VC with infrastructure intuition
You know your Snowflake from your Snowpark. You help founders scale infra and debug GTM. You know which AI dashboards are lipstick - and which are levers.
🧭 Step 4: Play Your Own Game
Don’t chase gold. Own a mine.
Own a vector
Deep in compliance tooling? Obsessed with RAG infra? Known for devtools with taste? Build there. You’ll see patterns no tourist can.
Design like you’re building a platform
Tools should become SDKs. Infra should become products. Think modular, exportable, forkable.
Avoid “AI for AI’s sake”
If the process isn’t broken, don’t fix it with a transformer. Fit matters. Feedback loops matter more.
✅ TL;DR
What to do | Why it matters |
|---|---|
Find leverage points | Proprietary data, embedded workflows, plug-and-play products |
Engineer repeatability | Logging, testing, automation = scale |
Pick your spot | Stack position = strategic leverage |
Stay disciplined | Own your niche, don’t follow hype |
✍️ Final Thought
Leverage is earned.
It doesn’t come from being early to a trend. It comes from going deep where others are shallow.
Look for the boring edge cases.
Listen for the expensive mistakes.
And build tools that compound - not just scale.
Signing off and signing zero checks,
SWEdonym