Case Study11 min read

Disciplined signals, governed swarms, and never your keys.

How FACTA embedded as fractional CAIO, CTO advisor and engineering lead to build Staxis — the non-custodial crypto-trading infrastructure behind 250 self-learning swarms, a 2.5%/3% circuit breaker and a five-stage governance pipeline.

Carolina Fogliato

Jul 19, 2026

PROJECT FACTS

CLIENT

Staxis (non-custodial crypto-trading infrastructure)

SECTOR

Crypto market intelligence · trading infrastructure

FACTA SERVICES

Fractional CAIO · CTO advisory · Full-stack engineering

TIMELINE

5 months (thesis → paper stage, 250 swarms live)

TEAM

3 FACTA leads + 4 client engineers

STATUS

Paper stage · Calibration Reports publishing monthly

Executive summary

Signal intelligence without custody — shipped, governed, and audited.

Staxis came to FACTA with an uncomfortable thesis for crypto: every existing option — pooled funds, retail bots, custodial wallets — asks you to give up your keys, your capital, or eighteen months of engineering time. The product they wanted to build would separate signal intelligence from execution custody: 250 self-learning swarms surface what to trade, when, and where, and the user acts on their own exchange. The keys never leave the wallet.

FACTA embedded across three planes — fractional CAIO direction, CTO advisory architecture, and engineering leadership — and shipped the platform to paper stage in five months: a typed HTTP API and SDK, a five-stage governance pipeline (paper → canary → live → degraded → retired), a 2.5%/3% daily circuit breaker with three-layer enforcement, hash-chained audit logs, and a monthly Calibration Report that publishes misses as well as hits. The first three Argonaut swarms are running; the remaining 247 are staged behind the same governance gate.

OUTCOMES

Time to paper stage

5 months

Thesis → 250 swarms architected, 3 in paper

Swarms governed

250

Same 5-stage lifecycle, none skipped

Custody permissions

Trade-only

Withdrawal scope never requested

Calibration cadence

Monthly

Brier-scored, misses published

The client

Staxis — the discipline of a systematic strategy, without the custody.

Staxis is a crypto market-intelligence company built on a premise that contradicts the segment it sells into: a disciplined trader shouldn't have to choose between handing custody to a black box and becoming a full-time quant. The product is not a fund and not a wallet — it is the infrastructure between a user's custody and a systematic strategy: a swarm of typed strategies, a daily circuit breaker, a documented governance pipeline, and a proprietary market-intelligence backbone, exposed through a dashboard, an API, and an SDK.

When FACTA was engaged, Staxis had a founder, a research thesis, and the first three Argonaut strategy sketches (Trend, Mean Reversion, Events). What it did not have was a production architecture, a governance model that a regulated client could audit, or a market-intelligence layer that could feed 250 swarms without one becoming a silent single point of failure.

The challenge

Build governed swarm infrastructure a regulator could read.

The scope went beyond shipping a trading app. Staxis needed an architecture where custody, strategy, and governance were separable — and where each separation was verifiable from a permissions scope, a property test, or a governance log. Every swarm had to pass the same five-stage lifecycle, with promotion criteria documented and demotion automatic. And the platform had to publish what it got wrong, not just what it got right — a Calibration Report with Brier-scored predictions and the misses left in.

CONSTRAINTS

Non-custodial by construction

Staxis must never receive withdrawal permissions. Even under compromise, the worst case is bad trades, not lost funds — verifiable from the OAuth scope, not a promise.

Risk control, not risk guarantee

A 2.5%/3% daily circuit breaker with three-layer enforcement and property-tested invariants — disclosed as a control, not insured against losses exceeding 3%.

Auditable governance

Five ordered stages (paper → canary → live → degraded → retired), minimum 90 days per pre-live stage, Brier-score gating, Council approval, and automatic demotion on regression.

Honest calibration

Monthly Calibration Reports that publish misses as well as hits, with downloadable CSV data and no fabricated performance numbers.

Regulated surface

Geoblocking (New York / BitLicense), California DFAL gating, and a SOC 2 Type I (Year 1) → Type II (Year 2) roadmap — the architecture had to be audit-ready from day one.

FACTA's approach

One partner across direction, architecture and execution.

FACTA embedded as a single technical partner operating across three planes that, in a regulated fintech, would normally require three separate leadership hires. The same person who defined the governance policy reviewed the property tests that enforced it.

01

Direction — Fractional CAIO

  • ·Product thesis: signal intelligence without custody
  • ·Swarm roadmap and the 250-swarm naming system (Argonauts)
  • ·Pricing and tier policy (Free → Pro → Plus → Power → Performance → Enterprise)
  • ·Calibration Report as a trust signal, not a marketing artifact

02

Architecture — CTO advisory

  • ·Non-custodial permission-scope design
  • ·Five-stage governance lifecycle and promotion/demotion triggers
  • ·2.5%/3% circuit-breaker invariants and the three enforcement layers
  • ·Market-intelligence backbone architecture (on-chain · market · macro)

03

Delivery — Engineering leadership

  • ·Led the team building the dashboard, typed HTTP API and SDK
  • ·Implemented hash-chained audit logs and the Calibration Report pipeline
  • ·Wired the Brier-score and drawdown gates into each stage transition
  • ·Set up geoblocking, SOC 2 evidence collection and the status page

The technical solution

Staxis: five properties, one set of invariants, three surfaces.

The final architecture organizes Staxis into five architectural properties that share one control plane. The dashboard, the API, and the SDK all run on the same non-custodial guarantee, the same circuit breaker, and the same audit log — nothing is hidden behind a second engine.

MODULES

01

SWARM RUNTIME

250 self-learning agent swarms, each running one documented strategy with deterministic risk gating.

02

CIRCUIT BREAKER

2.5% / 3% daily loss control with three-layer enforcement and property-tested invariants.

03

GOVERNANCE PIPELINE

Five ordered stages — paper → canary → live → degraded → retired — with minimum 90 days per pre-live stage and automatic demotion.

04

MARKET INTELLIGENCE

Proprietary backbone fusing on-chain, market and macro signals into the inputs each swarm reads.

05

CALIBRATION REPORT

Monthly Brier-scored publication of predictions, realized outcomes, misses, and what changed.

06

SURFACES

Dashboard, typed HTTP API and typed SDK — one set of invariants, three integration paths.

STACK

Backend
Typed service layer · hash-chained audit log · property-tested invariants
Surfaces
Dashboard (web + mobile) · typed HTTP API · typed SDK for client stacks
Intelligence
On-chain · market · macro signal fusion feeding 250 swarms
Risk
2.5%/3% daily circuit breaker · three-layer enforcement · Brier-score gating per stage
Governance
5-stage lifecycle · 90-day minimums · Council approval · automatic demotion
Compliance
Geoblocking (NY/CA) · SOC 2 Type I → II roadmap · 7-year audit retention

Non-custodial · Governed · Audit-grade

REPOSITORY

github.com/facta/staxis

Results

Five months. Paper stage. Audit-grade transparency.

Staxis reached paper stage on schedule, with the first three Argonaut swarms running against real market data and the remaining 247 architected behind the same governance gate. The first Calibration Report was published with downloadable CSV data, Brier-scored predictions, and the misses left in — no fabricated performance numbers, no curated success stories.

The architecture shipped audit-ready: geoblocking enforced for New York, California gated pending DFAL compliance, and SOC 2 Type I evidence collection running from day one. The non-custodial guarantee is verifiable from the permissions scope — withdrawal is not requested — and the circuit breaker is a documented control, not a guarantee of maximum loss.

METRICS

Time to paper stage
5 months
Thesis → 3 Argonaut swarms in paper
Swarms under governance
250
Same 5-stage lifecycle, no stage skipped
Custody permissions
Trade-only
Withdrawal scope never requested
Calibration Reports
Monthly
Brier-scored, misses published, CSV downloadable
Surfaces shipped
3
Dashboard + typed API + typed SDK
Audit readiness
SOC 2 Type I
Type II scheduled for Year 2

“FACTA was the only partner who could write the governance policy in the morning and the property test that enforces it in the afternoon. That's what let us ship 250 swarms under one auditable lifecycle instead of a black box.”

Founder & CEO

Staxis

Replicable lessons

What makes a regulated crypto product shippable in five months.

Non-custodial is a permissions scope, not a slogan.

The guarantee is verifiable from the OAuth scope the user grants — withdrawal is never requested. That moves trust from a marketing page into something a regulator can read.

A circuit breaker is a control, not a guarantee.

The 2.5%/3% breaker has three-layer enforcement and property-tested invariants, and the failure modes — market gaps, slippage, outages — are disclosed in full. Honesty about the control's limits is what makes it credible.

Governance is a lifecycle, not a stage.

Five ordered stages with minimum 90-day durations and automatic demotion turns 'trust us' into 'read the log'. Every swarm, including roadmap ones, passes the same gate.

Publish the misses.

A Calibration Report that includes what the swarms got wrong is more persuasive than any backtested equity curve. Audit-grade transparency is the moat in a segment that runs on opacity.

About FACTA

Fractional AI leadership with production-grade engineering.

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 — including the property tests, the audit logs, and the published misses.

We combine ten years of experience in AI architecture with operational expertise in multi-agent systems, market intelligence, governance pipelines, 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, with a particular focus on regulated and audit-sensitive products.

SERVICES

01 · Fractional CAIO

C-suite-level AI direction. Product thesis, swarm roadmap, pricing policy, market positioning.

02 · Fractional CTO

Architecture, non-custodial design, governance lifecycle, risk controls, and compliance trade-offs.

03 · Full-stack engineering

Implementation with an in-house team: dashboard, typed API, typed SDK, audit logs, calibration pipeline.

04 · Strategic Advisory

Supporting founders on regulatory posture (BitLicense, DFAL, SOC 2), pricing transparency, and go-to-market for regulated crypto products.

Building something audit-sensitive?

If you're shipping a product where governance, risk controls, and published calibration matter as much as the signal, let's talk.

Book a 30-minute call

hola@facta.dev

NEWSLETTER

AI strategy and engineering insights for technical leaders — every Tuesday.

Join 2,500+ engineers and leaders getting practical AI implementation insights, including governance and calibration patterns.

Subscribe

Your Privacy Matters

We use cookies to enhance your experience, analyze traffic, and serve targeted ads.

By clicking "Accept All", you consent to all cookies. Cookie Policy