Field notes
The Moffhub blog
What we're building, what we're breaking, and what we've learned shipping software in East Africa.
One AI Core, Many Institutions: How We Built a Retrieval Engine You Can Trust
How Moffhub Solutions built a grounded, citation-first AI knowledge platform, from certified document ingestion to legal-grade hybrid search and human handoff.
Feature-based billing for African SaaS moffhub/billing v1
A Laravel billing engine built around how money actually moves in East and West Africa twelve payment providers, fallback chains, idempotent usage metering, multi-currency from day one. The thing Stripe Billing isn't.
Multilingual USSD that doesn't fall over moffhub/ussd
A USSD framework with translation keys, build-time length checking across every locale, intelligent session recovery, circuit breakers on external API calls, encrypted PIN fields and type-safe menu enums. Everything you wish you didn't have to write yourself.
Database-driven state machines for Laravel revenue systems
Why we extracted moffhub/flow from production county-revenue work and the pattern of approval gates, role guards and auditable transitions that comes out of it. Post 1 of the moffhub/flow series.
USSD still matters and we built a PHP toolkit for it
Smartphones haven't replaced USSD in East Africa. moffhub/ussd is the session-aware, testable PHP toolkit we use to ship menu apps that survive flaky 2G and Africa's Talking webhooks. This is post 1 of the moffhub/ussd series.
Four-eyes approvals as a Laravel trait moffhub/maker-checker
Drop a trait on a model and every save/delete becomes a pending approval request. Multi-role thresholds, user-specific approvals by email, payload-conditional rules, delegation with expiry, pessimistic locking, and a REST API for the approval inbox without writing any of it yourself.
Pair-programming with AI, for working engineers
AI as the partner you always wanted, but only if you keep doing the parts of programming that make you better. The patterns we use, the failure modes we have learned to spot, and the rules we put around the agent. Post 8 of the AI Foundations series.
AI in your daily workflow, for people who do not code
Concrete patterns for embedding AI into knowledge work. Drafting, summarising, researching, dictating, planning. The skill is recognising the next opportunity to apply the pattern. Post 7 of the AI Foundations series.
AI tools worth your time in 2026
An opinionated tour of the AI tools we actually keep open every day. What each is good at, where each falls down, and why a small toolkit beats a sprawling one. Post 6 of the AI Foundations series.
Inside an LLM, how language models actually work
A non-mathematical look at what is happening inside a large language model when you press send. Tokens, embeddings, attention, and why "predict the next word" turns out to be enough to write code and pass exams. Post 5 of the AI Foundations series.
Page 1 of 2 · 17 The Moffhub blog