Ethan Keiser

Coding bootcamps as a product largely collapsed in 2024–2025. The job market was the trigger, but the deeper issue was structural: many bootcamps optimized for enrollment instead of building defensible IP. I’ve been saying this for years. When I pushed bootcamps to focus less on volume and more on the engineering value they were creating, I caught plenty of heat for it. What felt obvious to me eventually became reality, and some still won’t acknowledge it. In tech, IP matters. It's the novel way of doing things captured in real code. It creates value even when customers slow down. It can make you an acquisition target when things go sideways. It also builds credibility with strong technical people, and that credibility compounds. Community and raw user counts can help, but they’re fragile. If the community fades, the value often goes with it. Too many bootcamps had little to no IP. They relied on generic or free curriculum, stitched together third-party tools, and called the glue a product. That works in a boom. It fails when conditions tighten. The silver lining is simple: admit it, learn, and build something real. Instead, I’m seeing defensiveness and deflection, and it’s making the industry’s decline sad to watch.

53

13 Comments

We secured over $400k ARR before writing a line of code. It was intentional. Finding PMF is incredibly hard because there’s no one playbook And the most important part of PMF is actually… the market. You cannot validate the market has a desperate, hair on fire, pain point based on discovery meetings. You have to validate with behavior. But once we saw customers were willing to spend 6-figures to solve a problem where the product didn’t exist yet… we knew we found a problem worth solving Here’s how I’m thinking about it now: 1. Market Validation: Validate you’re solving a problem worth solving with purchase behavior. Discovery calls aren’t validation. 2. Then - Product Market Fit: Validate by retention, not revenue. Identify the lead product usage indicators that validate someone is living and breathing your solution. 3. Then - Go-to-Market Fit: Validate unit economics. Break down the lead indicators of pipeline velocity and CAC. We have weekly measures around # of SQLs + cost / SQL One of the primary reasons startups don’t make it is scaling too early and too quickly. But most do not have a clear and measurable scorecard that tells them ‘when’ to scale and ‘how fast.’ I’m doing things differently this time around and posting publicly. Follow our journey here!

72

13 Comments

Build for scale. Don’t build at scale. One of the easiest ways to kill early momentum is trying to engineer for millions of users before you even have 100. Design your system to handle growth, but stay lean enough to move fast, iterate, and learn. Premature scale = wasted time, money, and complexity. Scale when it matters. Not before.

86

11 Comments

🔥 We split our engineering team before it broke… and that’s exactly why it worked. The strange part? Nothing looked broken. Shipping was solid. Velocity was green. No fires. No drama. But under the surface… things were getting heavy. More cross-dependencies. More “quick syncs” turning into >10 people meetings. More context switching. More moments where 1 change required 6 people to “stay in the loop.” Classic signs of a team that’s too big to stay fast, but not broken enough to force change. We didn’t wait for the slowdown. We evolved the structure before the bottleneck became visible. Because growing teams need growing structures. And every seasoned CTO will tell you the same thing: after ~8–12 engineers, coordination becomes the real constraint. 🚀 So we made the call: two focused squads. Prospect → Lead discovery + intake (data pipelines, enrichment, integrations, workflow automation) Engage → Outreach + follow-up (campaigns, inbox, dialer, analytics, identity, chrome extension) Clear domains. Clear ownership. No overlaps. Two proper two-pizza teams that can move fast without tripping over each other. And for everything that didn’t fit neatly into either domain, infra work, shared tools, odd cross-team tasks, we carved out separate capacity so it wouldn’t derail the core squads. That one choice eliminated a ton of hidden drag. ⚡ The impact was immediate. Context switching collapsed. Engineers stay in one world per day instead of juggling six. PRs fly through. Decisions now happen inside the squad, not across the org chart. Planning became simple. Each squad runs its own backlog, its own cadence, its own momentum. Ownership went up. Every engineer is getting to know their domain end-to-end, and it shows in the work. The vibe changed. Same speed, far less chaos. More flow. Less noise. Actual calm execution. A month in, the signal is unmistakable: Splitting didn’t slow us down, it kept us fast as we grew. 🧠 The real lesson? If you want engineering to stay fast, you can’t just add people. You have to reshape the system they operate in. Small, focused teams with real ownership will always outship a single bloated group. We didn’t wait for a crisis to restructure. And that one proactive move preserved our momentum when it mattered most.

136

9 Comments

How much money do mobile developers make today? Back when I dabbled with Kotlin and iOS, freelance mobile developers were charging north of $200/hr. I'm not sure about today, but AI is definitely changing things. You can now build cross-platform mobile applications using Emergent's AI platform. This is the first platform I've seen that's going after mobile with such force. They are also releasing a Pro mode, which will make the vibe-coding experience even more powerful: 1. You can update your system prompt 2. You can build custom agents Both Mobile App Development and Pro Mode are in a limited pre-launch phase. If you want to try before they close sign-ups, you should go to app.emergent.sh and sign up. Something to keep in mind: Emergent was trained from the ground up for building web and now mobile applications. That's the only way to make a platform that builds production-ready applications. Thanks to the Emergent team for collaborating with me on this post.

30

2 Comments

Today many will think I’m killing my company. From now on, at Pullpo, we’re allowing our clients to purchase the full Pullpo codebase. Every line of code. As a one-time purchase. Why we are doing this: DevEx platforms make a ton more sense when they’re fully customized to every aspect of a team: their workflows, their internal tools, their automations and the very specific metrics that matter to them. That’s the perfect DevEx platform. But as a SaaS, it’s impossible for us to deliver that level of customization for everyone. Every client has different pains and priorities, and as a company we can’t satisfy all of them quickly. Some requests are so specific that they don’t make sense to consider in a generic product. This is changing today. Just two months ago, it would have made zero sense for a small/medium team to spend precious engineering time maintaining and evolving its own DevEx platform. You’d rather ship product than build internal tooling. The cost, the complexity, and the ongoing maintenance burden were simply too high. But the equation has changed. Building in-house is getting faster and easier every month with AI. Integrations are simpler. Infrastructure is more plug-and-play. And more and more “new features” look less like multi-week projects and more like a well-structured prompt plus a thin layer of engineering. That’s why we’re providing teams with a SOC 2–audited, secure, optimized, production-proven base. So they can build their perfect DevEx platform on top of it. And we won’t just hand over the code and disappear. We’ll provide support, advice, and even forward-deployed engineers to help teams ship the custom features they need. Reach out if you want the perfect DevEx platform for your team.

40

5 Comments

Hilarious. Sloppy engineering patterns are on the rise. There’s a bifurcation between engineers using coding agents to amplify themselves and “vibe coders” going the other way. Simon Willison's recent blog proposed a new verb for the amplified engineer: "vibe engineering." His idea: there’s a big difference between sloppily prompting and accepting-all-changes for a clumsy prototype, and the intentional use of LLMs by seasoned professionals to accelerate production-quality code.  This is the part I really liked: "If you’re going to really exploit the capabilities of these new tools, you need to be operating at the top of your game. You’re not just responsible for writing the code—you’re researching approaches, deciding on high-level architecture, writing specifications, defining success criteria, designing agentic loops, planning QA, managing a growing army of weird digital interns who will absolutely cheat if you give them a chance, and spending so much time on code review." The semantics might be meaningless, the distinction isn’t. Amplify or reduce. Related thought that Yang and I have been tossing around: there’s room for the "Staff Engineer" level agent that issues tasks and coordinates “IC agents” - testing, integration, deployment, etc. What does the next generation of AI-enabled developers look like?

202

20 Comments

I built an iOS app with enterprise SSO and multiple integrations in under 6 hours this weekend. Not a toy. Not a tutorial project. A production-ready app based on existing business logic, with proper authentication. The tool: Anthropic's Opus 4.6, released last week. What's different? Tasks run longer. The model maintains context across complex, multi-step work without you babysitting every prompt. You describe intent, it executes. Six hours instead of six weeks. And here's what the "AI is killing software" crowd is missing: This doesn't replace developers. It compresses development. Faster builds don't mean fewer people. They mean more shipped. More experimentation. More ambitious roadmaps that were previously "too expensive to try." The $1 trillion selloff in software stocks this week was investors panicking that AI makes software obsolete. They've got it backwards. AI makes software easier to build. That means more of it, not less. The real gap isn't "can AI code?" — that question is settled. The gap is: can your team turn AI output into production? That's the skill that matters now. And most orgs don't have it yet.

47

19 Comments

When Claude Opus 4.5 generates an Artifact, it doesn't just render a React app. It incorporates the business, files the LLC paperwork in Delaware, and closes a Series A funding round before the UI finishes rendering.

182

19 Comments

🚨Payment folks: you're going to like this one. Just launched Fee Sponsor for Solana at Dfns. We built this after seeing payment companies struggle with a major UX hurdle: users needing SOL just to move USDC. The feature removes this friction entirely. Imagine a user on Solana who holds only USDC — no SOL: They can now transfer those assets without ever touching SOL 🔥 1. Your company sponsors the transaction fee from a separate wallet through our API 2. The transfer completes smoothly, all SOL logistics invisible to the user For payment providers, this solves three critical problems: • No more abandonment during onboarding because users can't obtain native tokens • Simplified treasury management; fund one sponsor wallet instead of thousands of individual wallets • Cleaner UX that abstracts blockchain complexity from end users Every sponsored transaction maintains full audit trails, policy enforcement, and permission controls. We're starting with Solana, with EVM support soon now that Pectra is live. If you're running payment flows and fees are causing friction, drop me a message.

91

26 Comments