We ran a 40-person consultancy for nine years and shipped software across five countries. Then we watched what AI did to the SDLC — and made a decision most agencies are still avoiding. We didn’t pivot. We started over. Smaller, faster, and deliberately so. This is what we built instead.
In 2022 we wrote an internal memo: AI will make traditional software consulting obsolete by 2030. It happened in 2026.
Beta Launch was Janaka’s first company. Started in Melbourne. Grew to 40 people and operated across the Netherlands, Australia, Oman, and the USA. The Colombo Stock Exchange. USAID CATALYZE. Takso. HAL Capital. 120+ products. A decade of evidence that the old model worked.
We were good at it. That’s not why we closed it.
We closed it because the model was designed for a world where the bottleneck was access to skilled engineers. That world is gone. The bottleneck now is thinking — clear product thinking, tight specs, systems-level architecture. Implementation is increasingly automated. Access to engineering talent is no longer the constraint.
A 40-person company with management overhead and a pipeline optimised for the old world is the wrong shape for this one. A two-founder studio with deep product craft, a rigorous spec-first process, and AI-native tools that multiply output is the right one.
We didn’t downsize. We redesigned.
Amazon’s two-pizza rule says: if you can’t feed a build team with two pizzas, the team is too big. We took that further.
Three people is enough for any product — if each of them can own a full surface end-to-end. Not three specialists waiting for handoffs. Three generalists with depth. When the spec is tight and the tools are right, team size stops being the constraint. Complexity is.
This is why we cap engagements at three people. Not because we’re small — because we’ve found that three people with the right process outship fifteen people with the wrong one, every time.
Pizza-team software isn’t a compromise. It’s the upgrade.
When the spec is tight and the tools are right, team size stops being the constraint.
Founded Beta Launch in 2017 and ran it for nine years across five countries — USAID, the Colombo Stock Exchange, HAL Capital, and 120+ others. Hired, fired, pitched, and delivered. Watched the SDLC die in real time and chose to act on it before the industry caught up.
At specshop.dev, Janaka leads strategy, client relationships, product thinking, and spec2web’s direction. He’s the person who will be on your discovery call. He’s also the person thinking publicly about what AI actually means for product teams, software delivery, and the people who build both.
Joined Beta Launch as CTO and eventually ran all of engineering. Has architected systems for financial services, government, healthcare, and consumer apps. The person who gets called when the architecture needs to hold at scale.
At specshop.dev, Tiran owns the technical architecture on every build engagement, the spec2web toolchain, and the boundary system that makes AI-agent builds safe to ship. The person most likely to catch the thing everyone else missed.
These aren’t values on a wall. They’re constraints we built the process around because we learned them the hard way across 9 years and 120+ products.
We don’t start with wireframes or mood boards. We start with who does what, in what order, and what they see at every step. The screens are implied by the journeys — not designed separately from the logic. This is why our design phase is not a phase. It’s a byproduct of clear product thinking.
Nothing gets built until the spec is approved. This feels slow. It isn’t. Every hour of spec work saves three hours of code rework. If the spec is wrong, code makes it more wrong, faster.
AI agents handle implementation. Humans sign off on every stage. Not because AI is untrustworthy, but because the thing that needs to be right is understanding — not just correctness.
A three-person team where each person can own a full product surface is worth more than a ten-person team where everyone is a specialist waiting for a handoff. Generalist depth at pizza-team scale isn’t a constraint — it’s how we outship studios ten times our size. The bottleneck was never headcount. It was clarity.
We tell you on the discovery call what the real cost and timeline looks like. If it doesn’t fit, we’d rather lose the engagement than set up a project to fail.
We build in public. Every decision, failure, and course correction gets published when it happens — not after it’s been polished into a case study. Honesty compounds.
We’re not framework missionaries. The stack is chosen per engagement based on what the product actually needs — and what the client’s team can maintain after handoff.
We take on a limited number of engagements at a time — not because we’re precious, but because we can only do our best work on the things we’re fully inside. We tell you on the discovery call whether it’s a fit. If it isn’t, we’ll say so and point you in the right direction.
30 minutes. No pitch. Just clarity.