What it cost us to learn it
Worked examples on the agent stack and on the cost lever, run against ourselves first.
We publish what we know because we ran it on our own work before we offered it to anyone else. Each case study reports the receipts, the wiring, and the honest limits.
Read the index, pick the one closest to your question, read the receipts.
Public case studies
A short list of the work we have actually done
A short list of the work we have actually done, what it cost us to learn it, and what walking the same path with you would look like in scope. We post case studies after the engagement, not before. No promotional placeholders.
Substrate · Lactate
Lactate, the early warning signal
The walkthrough on continuous lactate monitoring: how the signal works, where it fits as an early warning of metabolic drift, what is mechanistically defensible, what is still in front of us, and where GNL will and will not go on this conversation. Anchored by 33 structured assessments across the GNL evidence base; longitudinal validation against HbA1c progression pending.
Cost lever · Claude Code
The £1,940 month, and the wiring that closes it
Eleven days into May 2026, our own Anthropic API spend climbed to £1,940. We caught it, fixed it overnight, and projected the forward saving. The case study walks every receipt: the three flags that should have caught it earlier, the wiring that closes the leak, the projected forward saving. Light Touch to Strong Signal in scope.
Architecture · RAG
Wiring an ecosystem-wide RAG mirror into Claude Code
Clarity by subtraction in three hours of overnight work. The infrastructure was already on disk for three days; the Claude Code sessions just did not know to use it. The wire-up: a single import, a CLAUDE.md HARD RULE, a slash command. Read-only mirror, zero new tokens, ecosystem-wide retrieval. The Via Negativa worked example.
Agent stack · Internal
The agents cycle on ourselves
What happened when we ran the seven-agent stack on Via Negativa as if we were a client of our own consultancy. Thirteen public pages and a new case study, audited end to end against locked HARD RULES. The receipts, the ROI projections, the honest limits: which agents earned their keep, which got Refer-back-up, what we would change before running it on a paying client.
Engagement scope
Indicative engagement scope
Each public case study is the worked example: Light Touch to Strong Signal in shape, scoped and priced in conversation. We do not publish a price for the public case studies because the figure is the wrong anchor; the scope is the right one. Send us the question, we will tell you which tier fits.
The Via Negativa worked-example framing: we publish what we know because we ran it on our own work before we offered it to anyone else. The receipts, the wiring, the honest limits, in that order. Anything we have not run on ourselves yet stays out of the index.
Private engagements
What we will not post
Partner engagements (device manufacturers, health systems, named clinicians) sit behind a short login, shared 1:1 under NDA. The public case studies are the ones where the story is ours to tell, end to end, with our own receipts.
Talk to us. One inbox: john@theglucoseneverlies.com