Insights
Notes from the build
Engineering and AI delivery notes from the ToolTwist team how we scope, build, and ship production software.
APIs, Microservices & MCP: How Software Talks to Software — and Now to AI Agents
A plain-language guide to the modern integration terminology — what an MCP server actually is, why it sits on top of your existing APIs, and how to explain all three in one sentence.
Read insight2026
17 posts- Methodology
You Do Not Need Spec Kit If You Are Already Using BMAD
Why bmad-spec covers spec-driven development and why running both tools creates more problems than it solves
- Engineering
Set Up Your Claude Code Status Line (Almost Nobody Does This)
A 30-second tweak that keeps your model, remaining context, rate-limit usage, working directory, and git branch on screen the whole time. It's the most useful Claude Code setting nobody talks about.
- Methodology
Run /insights, Then Let Your Most Expensive Model Grade You
The highest-leverage thing your best model can do isn't write another function — it's look at how you work. Here's a two-command habit: get Claude to report on your own sessions, then hand that report to Fable 5 for a brutal critique.
- AI delivery
Your AI Just Built You an App. It Also Built You a Liability.
A new Checkmarx report finds the more code AI writes, the more vulnerable it ships — and even mature security teams can't keep up. Here's what that means if you're building with AI.
- Methodology
BMAD vs. OpenSpec vs. Spec Kit: A CxO's Field Guide to Spec-Driven AI Development
An independent look at the three spec-driven frameworks dominating AI development in 2026 — and how a CxO should match each to project risk, team size, and compliance exposure.
- AI delivery
The AI Replacement Theatre
The headlines say AI is replacing workers. The operational data says something different. What's actually happening at Spotify, Uber, IBM, Salesforce, Google, and Meta.
- Methodology
Why Your BMAD Workflow Needs TEA And What Happens When It Does Not
The case for adding a dedicated test architecture layer to your AI-assisted delivery pipeline
- Methodology
Whether You Use BMAD, Spec Kit, or Just Vibes, your AI-Generated Tests Have the Same Problem
Why test architecture is orthogonal to implementation methodology and what to do about it
- Engineering
Why Your Test Coverage Number Is Lying to You
Why raw test coverage leaves brownfield systems exposed—and how risk-weighted gap intelligence helps teams close the gaps that matter.
- AI delivery
Job Genie: from AI market research to a working MVP
Using AI to validate the market, choose an architecture, integrate six gateways and payment services, and ship a Flutter MVP for Job Genie an SMS-blaster for last-minute temp recruitment.
- Engineering
Spread-em: designing a distributed file store with AI
How we used Claude to research and build Spread-em a bit-torrent-class distributed storage system for serverless social apps to 60% product readiness in five days.
- Methodology
How We Ship AI-Generated Code That Survives Production
A practical guide for SMB owners, founders, and product leaders who want the speed of AI without the technical debt ToolTwist's Spec-Driven Development framework and its six production guardrails.
- Migrations
Old-gauges: migrating frameworks with AI
Rebuilding Old-gauges an ESP32 IoT SaaS from a messy AWS Amplify/Lambda/DynamoDB stack to Next.js + MySQL, with Claude doing the framework migration in 2 days.
- Company
Engineered by experts, accelerated by AI
How ToolTwist pairs senior engineers with AI tooling to scope, build, and ship production software for enterprises, ISVs, and startups.
- Migrations
ToolTwist website Designer, rebuilt with AI
Reproducing our 20-year-old Java/JSP drag-and-drop website builder hundreds of widget types, thousands of pages in Next.js with an embedded MCP server, to 40% product readiness in 5 days.
- IoT & Firmware
Internet of Things: firmware generated with zero human coding
Using AI to write ESP32 firmware for dual wrist-mounted motion-capture watches Bluetooth sync, haptics, high-rate IMU and GPS to a working MVP in 3 days, with no human coding.
- AI delivery
Motorcycle Motion analytics: replicating a PhD project with AI
Reproducing a 2012 motion-analysis tool originally a PhD's worth of mathematics and months of research in a few days with Claude, including auto-written import adaptors for undocumented binary formats.