Notes from the build.
Writing from real projects, not thought leadership. The platform work behind this practice happens in the open, and these posts document what shipped, what broke, and the engineering calls in between.
SPEAKMAN.AI v1.3.0: server-side rendering, Copilot Studio, and a signed build
Document rendering moves onto the MCP server, dropping the local script and Node dependency generate-sad and generate-compliance-report used to need. That's the change that let generate-sad run end-to-end inside Microsoft Copilot Studio for the first time. The Windows build is signed, and four engine bugs are fixed.
Read the post →speakmanai-cc: now running in Microsoft Copilot Studio
generate-sad and generate-compliance-report ran end-to-end inside a live Copilot Studio agent for the first time, after their rendering step moved onto the MCP server. All 17 skills now package into a single Copilot Studio upload bundle too.
Read the post →SPEAKMAN.AI v1.2.0: a real human review gate, delegate mode, and UX design
A human-in-the-loop revision gate alongside the automated validator, a delegate mode that hands work to your connected coding agent, a tool that revises a finished session without starting over, and a workflow that turns architecture into clickable mockups.
Read the post →speakmanai-cc: two new SDLC phases
Idea intake and UX design join the pipeline, bringing the Claude Code skill count to 17. The whole set now ships to Claude Desktop in a single upload, no per-skill setup required.
Read the post →We built a pentest platform. Then we broke into it.
An AI orchestrator runs real recon, scanning, and exploitation tools against an approved scope, with four human sign-off gates in the middle. We validated it by pointing it at its own companion app: ten findings, four critical, every one expected.
Read the post →The SDLC as Claude Code skills
Fourteen markdown files take a project from a business description to a pentested, production-ready codebase. One command runs the sequence. Every human checkpoint still needs a human.
Read the post →SPEAKMAN.AI v1.1.0: a middle tier, Vertex AI, and requirements engineering
The free local engine gains a standard model tier, native Vertex AI configuration, and a workflow that turns a business description into implementation-ready technical contracts. Plus the bugs that only months of real use surface.
Read the post →Subscribe via RSS · Or talk to us about the work behind these posts.