PROJECT FACTS
CLIENT
Ctx0 (organizational AI infrastructure)
SECTOR
AI platform · self-hosted, source-available
FACTA SERVICES
Fractional CAIO · CTO advisory · Platform engineering
TIMELINE
5 months (thesis → 6 live modules)
TEAM
3 from FACTA · 4 from the client
STATUS
In production · 6 native modules live
Executive summary
AI infrastructure that compounds instead of forgetting.
Ctx0 came to FACTA with a thesis that the AI ecosystem refuses to confront: individual assistants remember one chat, agent frameworks remember tasks, but the organization keeps forgetting — 42% of institutional knowledge walks out the door with every departure, and every prompt starts from zero, costing 3–5x in tokens. The vision was a single platform — collective memory, a governed AI workforce, and proof of ROI — built as six native modules sharing one Platform → Organization → Project hierarchy.
FACTA operated as end-to-end technical partner: fractional CAIO direction, CTO advisory on architecture and trade-offs, and platform engineering on the modules that today make up Ctx0 — HIVE, FORGE, SWARM, ORACLE, NEST and Annie. The platform shipped to production in five months with all six modules live, source-available and self-hostable, with HIVE cutting token spend 40–60% on the FACTA codebase and ORACLE measuring cost per approved output as the only ROI metric a closed loop can compute.
OUTCOMES
Time-to-production
5 months
Thesis → 6 live native modules
Token reduction (HIVE)
40–60%
Measured on FACTA codebase
Outputs unreviewed
0
FORGE approval gate enforced
Native modules live
6
HIVE · FORGE · SWARM · ORACLE · NEST · Annie
The client
Ctx0 — the platform thesis the AI ecosystem refused to confront.
Ctx0 is an AI infrastructure company founded on an uncomfortable premise: companies have total visibility into cloud spend and zero visibility into AI spend. Prompts, outputs and critical decisions pass through third-party SaaS with no organizational memory, no per-project audit, and no cost control. Ctx0 was conceived to close that gap with a closed loop — memory in HIVE, planning in FORGE, execution in SWARM, human approval at the gate, measurement in ORACLE — so the same question the board keeps asking could finally be answered: is the AI actually paying off?
The challenge
One closed loop across six modules, not a pile of services.
The ambition wasn't a chatbot or another agent framework. It was six modules sharing a single control plane and a single memory graph, each one live and self-hostable, with the entire loop auditable from a prompt to a human-approved output. The hard part wasn't any single module — it was the closed loop: HIVE had to brief SWARM, SWARM had to route through FORGE, FORGE had to write approved work back into HIVE, and ORACLE had to read every module directly so nothing self-reported.
CONSTRAINTS
Memory that compounds
Knowledge had to survive departures and earn its reach — proven patterns promote from Cell to Comb to Colony, superseded nodes get pruned, idle nodes flagged stale at 90 days. Not a wiki — a living graph of the why.
Governed execution, not chatbots
Specialist crews per role — analyst, architect, implementer, reviewer — each briefed from HIVE memory, with every output routed through a human approval gate before it ships or enters memory.
Measured ROI, not vibes
Cost per approved output — total LLM spend divided by outputs a human actually approved — sliced per member, model, project, task type and time period. The metric only a closed loop can compute.
Self-hosted by default
AES-256-GCM at rest, RBAC, BYOK, exportable audit logs. Source-available and deployable on cloud, on-prem or air-gapped — no vendor lock-in.
FACTA's approach
Triple role across direction, architecture and execution.
FACTA came in as a technical partner — not an hours vendor — operating simultaneously across three planes that would normally require three separate full-time hires:
01
Direction — Fractional CAIO
- ·Product thesis and the closed-loop narrative
- ·Module roadmap and the six-module scope
- ·The 'cost per approved output' positioning
- ·Investor and partner dialogue
02
Architecture — CTO advisory
- ·Platform → Organization → Project hierarchy
- ·Memory graph design (Cell, Comb, Colony)
- ·Approval-gate and governance model
- ·Stack decisions and buy-vs-build trade-offs
03
Delivery — Platform engineering
- ·Led the team building HIVE's GraphRAG hybrid retrieval
- ·Shipped SWARM's crew registry and FORGE's approval gateway
- ·Stood up ORACLE's telemetry and killer-metric pipeline
- ·Owned the documentation and brand catalog as source of truth
The technical solution
Six modules. One closed loop. Memory that compounds.
Ctx0's final architecture organizes AI operations into six native modules sharing one Platform → Organization → Project hierarchy. Every action — a memory write, a crew dispatch, an approval verdict — feeds the same loop, so ORACLE reads the whole loop directly and nothing self-reports.
MODULES
HIVE
Collective memory. Three layers — Cell, Comb, Colony — with GraphRAG hybrid retrieval. Repositories become memory banks; every approved output compounds.
FORGE
Specs from real code, with a human approval gateway. Nothing ships, nothing is remembered, until a human approves it.
SWARM
Specialist crews per role — analyst, architect, implementer, reviewer — briefed from HIVE memory. Every output scored 0–100 before a human sees it.
ORACLE
Cost per approved output — total LLM spend divided by human-approved outputs. The killer metric, sliced five ways, recomputed hourly.
NEST
Self-hosted AI workforce hub. Models, agents, memory and traceability under one operational plane — cloud, on-prem or air-gapped.
ANNIE
The company's Jarvis — voice in, answers with sources. Surfaced for stakeholders who don't live in the admin.
STACK
Results
Five months. Six modules. A loop that compounds.
Ctx0 shipped to production on schedule with all six native modules live. The closed loop held: HIVE reduced token spend 40–60% on the FACTA codebase, FORGE enforced zero unreviewed outputs, and ORACLE produced the metric the platform was built to answer — cost per approved output, per member, per model, per project. The platform's own documentation became the source of truth for module names, taglines and routes, so marketing and engineering stopped disagreeing about what the product was.
METRICS
“Every other partner wanted to sell us a chatbot or an agent framework. FACTA was the one that understood the moat isn't the model — it's the memory that compounds and the metric a closed loop can compute. Five months later we had six modules in production and an answer to the board's question.”
Founder & CEO
Ctx0
Replicable lessons
What makes the loop hold.
The moat is memory, not the model.
Models are a commodity. The compounding graph — every approved output written back into HIVE — is the asset competitors can't close once it's running.
Governance is an architecture decision, not a policy document.
FORGE's approval gate is a runtime boundary, not a checklist. Zero outputs ship unreviewed because nothing can — the loop doesn't allow it.
Cost per approved output is the only ROI metric that survives a board.
Token counts and approval rates are ingredients. The metric that closed the loop — total spend divided by human-approved outputs — is what made ORACLE unkillable.
Make documentation the source of truth, not a deliverable.
Module names, taglines and routes all resolve from one brand catalog. Marketing and engineering stopped disagreeing about what the product was, because they read the same file.
About FACTA
Fractional AI leadership with real technical execution.
FACTA is a consultancy specialized in fractional Chief AI Officer and CTO roles for companies that need to lead the transition to AI without hiring a full-time executive team. Unlike generalist consultants who deliver PowerPoint, FACTA delivers systems in production.
We combine ten years of experience in AI architecture with operational expertise in multi-agent systems, Graph RAG, A2A, and a modern stack (Rust, TypeScript, cloud and on-prem infrastructure). We work with post-seed startups through Series B scale-ups across Spanish- and English-speaking regions.
SERVICES
01 · Fractional CAIO
C-suite-level AI direction. Product thesis, roadmap, investor relations, market positioning.
02 · Fractional CTO
Architecture, stack decisions, technical hiring, engineering oversight, governance and platform trade-offs.
03 · Platform engineering
Implementation with an in-house team: memory graphs, agent swarms, approval gates, telemetry, self-hosted deployment.
04 · Strategic Advisory
Supporting founders on fundraising, GTM, strategic partnerships and regulatory compliance.
A next case?
If you're building a platform that needs CAIO, CTO and engineering in a single partner, let's talk.
Book a 30-minute callhola@facta.dev