From our SaaS audits
What we find inside SaaS companies
Patterns from the technical due diligences we run for investors and boards.
AI
Pairing with a yes-machine
Pairing with an AI feels like pair programming. It isn't. There's one prior in the room, and it's trained to agree. This post makes the case for mob programming and spec-driven development as the structural fix.
The AI skills gap
In 40+ interviews, senior engineers from major banks and consultancies showed strong backgrounds but little real AI fluency. No RAG, no agent frameworks. The gap isn't about skill, it's about exposure.
Hermes: the agent that doesn't quit when you close your laptop
Hermes is an open-source AI agent that runs on a server, remembers across sessions, and builds reusable skills over time. The shift it represents: AI moving from something you summon to something that runs.
Building a customer support AI agent that learns before it speaks
A customer support AI agent built in stages: shadow mode first, internal notes second, auto-send only after the data earns it. A walkthrough of the architecture, the knowledge base design, and the lessons that held up.
LLMs everywhere, even in cars
LLMs are no longer a tab you open. They're the interface layer between intent and every system underneath. This post maps what ambient AI, edge inference, and agent-as-infrastructure mean for how you design modern software.
The end of the all-you-can-eat buffet
The all-you-can-eat era of AI is ending. Compute constraints, heavier models, and a fully hooked user base are pushing providers toward pay-as-you-go. That shift will force better choices, smaller models, and fiercer competition between tools.
ElevenLabs: voice cloning, agents, and what they mean for your product
Voice is where AI product differentiation is heading. This post walks through ElevenLabs voice cloning and conversational agents in enough detail to evaluate whether the technology is ready for your use case.
"Good news, I built it in Lovable.": an engineer's guide to surviving that sentence
Business users love Lovable. Engineers tend to panic. A real-world case study of how to wrap an AI builder in guardrails so non-technical teams can move fast without quietly rewriting the systems that give your product its edge.
Python as the new Latin
I used to teach people to code. And looking back, I was teaching students to write it by hand while the tools that write it for them were getting better every single month. So what should a coding classroom actually look like now?
Customer support in the AI era
Most AI-powered customer support is optimised for deflection, not resolution. The problem isn’t bad agents, it’s architecture: no shared context, no real permissions, no escalation path that works.
Mental capacity is a bottleneck
AI removes bottlenecks until it reaches the one that doesn’t move: human cognition. The faster AI makes your system, the more your team’s mental capacity becomes the constraint. You can’t add more of it.
Taste is the moat
When AI closes the execution gap, taste becomes the differentiator. Curation, judgement, and the willingness to say “not this” compound over time in ways that models can’t replicate.
Is the AI agent frenzy similar to the mobile app hype of the early 2010s?
In 2010, every business convinced itself it needed a mobile app. Fast forward to 2025, and the script is identical, just with AI replacing mobile as the technology everyone insists they can't afford to be without.
The authenticity tax
AI-generated content has made polished writing look suspicious. The deeper cost is the thinking you skip when you outsource the words that define your position.
Your limit will reset at 12pm
Half of today's AI best practices are coping mechanisms for temporary scarcity, not timeless engineering insights. Geoffrey Dhuyvetters traces the arc from SMS bundles to token limits, and argues the price curve only goes one direction.