Shipping a SaaS against a regulatory deadline, solo

LoadSnap is the waste-compliance SaaS I'm racing against DEFRA's October 2026 digital waste-tracking mandate. Eight months out, in private beta, with three customers onboarded. This is the build log.

The bet

Regulator hands the market a deadline. Incumbents drag. New entrants who ship early win share. Standard playbook.

The variables that matter for whether the playbook works this time:

1. The deadline is real and won't slip. (DEFRA confirmed in their February 2026 update: October 2026 stands.)
2. The incumbent vendors are slow. (Confirmed by talking to four operators. All three current providers are PDF-based and have been "developing the digital version" for two years.)
3. The audience is reachable. (Yes. Trade associations, waste-management forums, and Companies House data. About 12,000 operators in scope in the UK.)
4. A small team can ship a compliant product faster than a large one. (This is the bet. Eight months in, I think the answer is yes, partly because of Claude.)

The build approach

Solo. Claude Code as daily collaborator. Three custom agents inside the product. The rest of the stack is boring on purpose: Next.js, Supabase, Stripe Connect, Upstash, Resend, Groq for the latency-sensitive paths.

Three months of build, two months of compliance work with DEFRA's reference API, one month of customer development, two months of polish and bugfix.

What's working

Claude as architecture partner. Before writing the receipt-issuer agent, I spent three days with Claude going through the DEFRA schema field by field, asking it to spot ambiguities. It caught two that the reference docs glossed over. The compliance check would have failed in production on both.

The single-binary mental model. LoadSnap installs as one thing. Operator signs up, photographs a load, drives away. No five-tool stack to learn. The competition is building enterprise-style admin consoles. I'm building something that looks like a camera app with paperwork attached.

Stripe Connect. The waste industry pays differently than most. Sometimes the producer pays the carrier, sometimes the disposal site reverse-charges. Stripe Connect handles the money-routing complexity without me building it.

What's not

The "demo to deal" gap is bigger than I thought. Operators love the demo. They sign up. They use it for a week. Then real life intervenes: a regulator inspection, a busy week, a driver leaving. The product needs to survive those weeks without losing the user. I'm underweight on onboarding hygiene.

Schema drift. DEFRA's reference API is in beta and they update it without strong versioning. The compliance-checker agent had to learn to refresh its schema view daily and pin to the current version. Earlier in the build, a schema change broke validation silently and we shipped invalid receipts for half a day before I caught it.

Pricing. Per-receipt pricing punishes the high-volume operators who'd be my best users. Subscription pricing under-charges them and over-charges the small ones. I'm landing on a hybrid (subscription with included volume, overage at decreasing per-unit) but I'm not certain yet.

What I'd do differently

Two things.

Start with the on-call rota. When you're running a compliance product, every customer hit is a regulator-adjacent risk. I left on-call structure until month three. It should have been week one.

Validate the price before the build. I started building before I'd run pricing conversations with five operators. Reversed, the pricing model would have informed the architecture (specifically, where to spend on per-request inference vs. caching).

What's next

Two months of polish. Onboarding flow rebuild. A "compliance status" dashboard that an operator can hand to an inspector in fifteen seconds. Stripe Connect over to live on the first batch of paying customers.

If you're running a UK waste operation and you've got DEFRA's October date on your radar, come and try LoadSnap. The build is solo but the customer development isn't.

The wider point: shipping AI-native products against a deadline is faster than shipping non-AI-native ones, but only if the AI is inside the system, not bolted on. Claude inside the compliance loop changes what a small team can build. Claude reviewing the final UI is just nice-to-have.

Source: /content/articles/shipping-saas-against-a-regulatory-deadline.mdx

More writing → /writing. Got a brief? Get in touch.