Candor 6.0 rebuilds its gray scale in OKLCH: neutral, cool and warm ramps on one lightness scale, so changing temperature barely moves contrast. Tinted neutrals take their chroma from one curve, and warm surfaces come forward while cool ones recede.
In a Claude Code eval, AI agents told in CLAUDE.md to read a playbook opened it 0 times in 69 trials, while an Agent Skill loaded it every time. What went unread was the judgment: measuring how far apart two colors are, and sending brand changes to a person for sign-off. For human oversight of AI agents, the lesson is to check the transcripts, because a capable model's output looks finished either way.
The CLI-vs-MCP debate has covered tools the model already knows. It hasn't covered the harder case: agents needing domain expertise the model lacks. A domain CLI paired with a workflow playbook puts practitioner judgment — the kind the model can't generate on its own — into the agent's loop. The architecture is also a governance question: whose expertise travels through your tool.
WCAG's contrast formula was built for 1999 display calibration, not legibility. Hot pink on near-black scores 6.6:1 and passes — and isn't readable. OKCA closes the gap at the algorithm layer with chroma compression, polarity-aware scaling, and a zero-false-pass invariant proven by construction across sRGB.
A blog advocating for accessible design couldn't pass its own dark-mode contrast audit. The fix wasn't a CSS patch — it was a design system grounded in WCAG compliance, OKLCH color science, and inclusive design.
Standards calcify. Accessibility governance spent 25 years learning what happens when compliance diverges from outcomes — AI safety is setting up the same dynamic, with those lessons available in advance.
LLMs produce technically-valid outputs that fail to serve disabled users. Context is the intervention - examining prompt engineering patterns that shift burden from users to systems.
How WCAG's legal codification has created a paradox in digital accessibility governance, where organizations face risk when adopting better solutions—leaving disabled users underserved despite apparent compliance.
A case study documenting the evolution from 0/10 to 9.5/10 prompt engineering skill through the P+F+I+D framework, featuring real prompts, metrics, and the CLAUDE.md breakthrough.
How applying product development fundamentals transformed a workflow problem into an accessibility tool adopted across government agencies.
How listening, shared constraints, and translation between disciplines transforms the relationship between creative vision and technical reality.
How I used AI to build an offline transcription app that protects user privacy—and what it reveals about the democratization of software development.
Personal ethical stance.
Pawn002 blog Organizational disclaimer
Pawn002 blog accessibility policy
Pawn002 blog privacy policy
Pawn002 blog Ethical Guidelines for Content Creation and Updates
Pawn002 blog legal disclaimer
First post on new blog site.
