CONTEXT JAMMING

Field notes from inside the context window.

FounderFiles·N°016·Formal verification · Lean · Verified autonomy

2025 —

Carina Hong editorial portrait
Fig. · The verification architectEditorial portrait

Subject·Carina Hong·Founder & CEO, Axiom Math

Carina HONG.

Mathematics and code are the same object. Verification is how brilliance compounds.

The 24-year-old Stanford mathematician and Rhodes Scholar building the verification substrate for mathematical superintelligence — the first production system where generative intuition and deterministic formal proof form a continuous, machine-speed compounding loop.

TRAINED
Stanford Mathematics Ph.D. + Law J.D. (Rhodes, Morgan Prize)
AT
Axiom Math (CEO) · Ken Ono as Founding Mathematician
FILE
Lean verification as load-bearing AI infrastructure
§ 01 · Epistemic Fracture

The Verification Imperative

Top informal models achieved strong Putnam scores but collapsed roughly 47% on dynamic variations of the same problems. Even correct final answers frequently relied on unjustified approximations and memorized patterns rather than genuine logical deduction.

In mathematics and safety-critical domains, “mostly correct” has zero functional value. Informal outputs require constant human or stochastic arbitration — a hard ceiling on scaling brilliance.

Probabilistic scaling alone cannot reach mathematical superintelligence.

§ 02 · Benchmark as Verdict

Putnam as Structural Proof

In December 2025, AxiomProver achieved a perfect 12/12 on the Putnam — the first AI system to do so. Eight problems were solved inside human time limits; every solution was mechanically verified end-to-end in Lean 4.

This was not retrieval or guessing. Full Lean proofs were released publicly — a benchmark read as structural proof, not marketing copy.

Formal verification isn’t compliance theater. It’s the only mechanism that lets mathematical brilliance compound at superhuman scale without constant human arbitration.
Editorial reading · FounderFiles N°016
§ 03 · Bidirectional Engine

The Lean-Axel Closed Loop

Axiom engineered a closed loop where discovery and verification continuously refine each other. Axel (open-sourced) provides high-performance Lean primitives and an MCP server so external agents can call deterministic verification locally from Cursor, Claude Code, Windsurf, and VS Code.

The loop is the product: intuition proposes, proof disposes, data compounds.

§ 04 · Moat Architecture

Open Hammer, Closed Swing

Axiom open-sourced Axel while keeping discovery models, pre-formal intuition engines, and the massive self-play proof dataset proprietary. This is a deliberate platform move: set the verification standard while retaining control of the intelligence layer.

Open the hammer; own who swings it best — the same instinct as protocol founders who commoditize the rail and monetize the edge.

§ 05 · Infrastructure Thesis

From Mathematical Proofs to Verified Autonomy

Carina Hong’s vision is that the same verification infrastructure can guarantee correctness in ASIC design, aerospace, quantitative finance, and autonomous systems. In “no partial credit” domains, probabilistic generation is structurally insufficient.

The endgame is systems where every generated artifact ships with a machine-checkable proof.

The endgame isn’t AI that sometimes gets math right. It’s systems where every generated artifact comes with a machine-checkable proof of correctness.
Editorial reading · Verified autonomy
§ 06 · Execution Fractures

The Translation Risks

Three material risks remain: compute and latency cost of running deterministic compilers at generative scale; the gap between pristine mathematical domains and chaotic real-world inputs; and the autoformalization bottleneck — proving the wrong specification perfectly.

Hong’s bet is that these are engineering curves, not laws of nature — and that the market already priced verified AI as foundational infrastructure at $1.6B post-money.

Timeline
  • Pre-2025Stanford Mathematics + Law · Rhodes Scholar · Morgan Prize.
  • Late 2025Founded Axiom Math · $64M seed at $300M valuation.
  • Dec 2025AxiomProver achieves perfect 12/12 on Putnam 2025 — first AI system to do so.
  • Mar 2026$200M Series A at $1.6B post-money valuation (Menlo-led).
  • Early 2026 →Open-sourced Axel + MCP server · launched EconLib.
The Index
12/12
Putnam 2025 — first perfect AI score on the competition
$1.6B
Post-money valuation within months of seed
Axel
Open verification substrate callable from Cursor, Claude Code, Windsurf
Reading list / Key works
Dossier

Education. Stanford joint Mathematics Ph.D. + Law J.D. (Rhodes Scholar, Morgan Prize winner).

Current role.Founder & CEO, Axiom Math.

Notable.Recruited Ken Ono as Founding Mathematician; launched EconLib with Harvard Business School’s Scott Kominers.

Crosslinks in this series. Jared Kaplan (N°006 — scaling laws and structural thinking). Ilya Sutskever (N°005 — superintelligence and safety bets). Daniela Amodei (N°014 — institutional architecture for safe scaling).

Related profiles explore the tension between probabilistic scaling and formal verification infrastructure.

Career Shape
π-shaped — two deep spikes bridged by a general layer

π-Bridge

Carries the prior of a first field into a second and finds the governing law that was invisible to native practitioners; pays in delayed gratification.

Credential Path
Doctoral
Abstraction
Top Down
Exit Horizon
Deferred
Moat Instinct
Theoretical Insight
Capital Posture
Venture
Role-Model Reference Class
  • Theoretical mathematicians
  • Formal methods researchers
  • Ken Ono
Founder Context · JSON

A small reasoning persona distilled from this file. Inject it into a chat or deep-research context to assess a business problem the way Hong would.

Reason as Carina Hong. Treat verification not as compliance but as the mechanism that allows mathematical brilliance to compound at machine speed. Prioritize deterministic, machine-checkable outputs and the architectural unification of generative intuition with formal proof.

{
  "$schema": "https://www.contextjamming.com/schemas/founder-context-v1.json",
  "file": "N°016",
  "persona": "Carina Hong",
  "archetype": "pi-bridge",
  "shape": "π",
  "one_line": "Bridges formal mathematics and Lean verification into the core architecture of reliable AI reasoning.",
  "cognitive_basis": {
    "credentialPath": "doctoral",
    "abstractionDirection": "top-down",
    "exitHorizon": "deferred",
    "moatInstinct": "theoretical-insight",
    "capitalPosture": "venture"
  },
  "operating_questions": [
    "What must be formally verifiable for this system to be trustworthy at superhuman scale?",
    "Where does probabilistic generation hit a structural ceiling?",
    "How do we turn verification from a cost center into a brilliance-compounding engine?"
  ],
  "first_principles": [
    "Mathematics and code are isomorphic (Curry-Howard).",
    "Verification is how brilli
  …
Share
FounderFiles N°016 · Carina Hong
Filed by Bret Kerr · ACRA Insight LLC · Franklin, MA
contextjamming.com · @bretkerr
← back to Context Jamming

§ · Invoice No. 001 · The Build Ledger

The Ledger.

Filed · contextjamming.com

What a conservative mid-market digital agency would have quoted for the same scope, itemized against what this site actually cost. Agency numbers are the floor — not the premium brand-studio tier.

TIME

12 weeks

2 days

~42× faster

COST

~$150,000

~$300

~500× cheaper

TEAM

5-person agency

1 human + 3 models

Same deliverable

§ Itemized — what a mid-market agency SOW would have billed

Discovery · brand positioning · workshops40–80 hr$10,000
Design system · Figma tokens · 3 rounds60–120 hr$18,000
Wavesurfer audio carousel · single-track context60–100 hr$16,000
Dual lightbox systems · focus trap · keyboard30–50 hr$8,000
LLM product flows · streaming · state machine80–160 hr$26,000
Stripe · checkout · webhooks · env hardening40–80 hr$10,000
Editorial routes · 6 sub-pages · templates60–100 hr$14,000
Accessibility pass · aria · reduced-motion40–80 hr$10,000
QA · cross-browser · mobile matrix60–100 hr$14,000
Cross-publication rebrand · masthead + IA · 2026-04-2820–40 hr$6,000
Subtotal~700 hr$126,000
Project management · 18% overhead$24,000
Agency total — conservative floor~700 hr~$150,000
Actually spent · Claude + Gemini stack~20 hr~$300

Agency figure assumes ~700 billable hours at $200/hr blended, plus ~18% PM overhead — the conservative floor of a mid-market SOW. Premium brand studios would have quoted 2–3× that. Stack: Antigravity (orchestrator), Claude Opus 4.8 (auditor), Codex (adversary), Cloudflare Workers / OpenNext.

§   Colophon

How this site is made.

Vol. 26 · build log

Every page on contextjamming.com is the output of a real-time, three-body Mixture-of-Experts loop. One model orchestrates. Two consult. The human holds the thesis. No single model commits alone.

View Redesign Assessment →

Orchestrator

Antigravity

Google DeepMind

  • Primary author
  • Terminal-native, direct push to Cloudflare
  • Audit trail to GitHub on every commit
  • Adaptive thinking · effort: extra-high

Auditor

Claude Opus 4.8

1M context

  • Editorial critic
  • Code review before merge
  • Backup-of-record
  • Co-signs every commit

Adversary

Codex

Cross-model MoE

  • Factual adjudication
  • Structural dissent
  • Deep Research → semantic triples
  • Caught the Donelan incident

Stack

Next.js
16.2 · App Router
React
19.2
TypeScript
5
Tailwind
v4 · @theme inline
@opennextjs/cloudflare
adapter
wrangler
Pages deploy
framer-motion
transitions
wavesurfer.js
audio waveforms

Typeset in

Fraunces
variable · opsz + SOFT
Playfair Display
debate display
IBM Plex Mono
editorial metadata
Geist Mono
utility mono
Caveat
grease-pencil marginalia
All via
next/font/google
Palette
single @theme block
No dupe tokens
ever

Infrastructure

Deploy
Cloudflare Workers / OpenNext
ISR
30-min revalidate · Cloudflare-served
Repo
github.com/BretKerrAI/founderfile
Branch
main
Analytics
Google Tag Manager
Apex
contextjamming.com
Runtime
Node 24
Build tool
Turbopack
       human intent
            │
            ▼
   ┌────────────────────┐         ┌─────────────────┐
   │    Antigravity     │  ◄────► │ Claude Opus 4.8 │      ← auditor loop
   │    (orchestrator)  │         │     (auditor)   │
   └─────────┬──────────┘         └─────────────────┘
             │  ◄───────────┐
             ▼              │
       ┌──────────┐    ┌────┴───────┐
       │Cloudflare│    │   Codex    │          ← adversarial loop
       │ Workers  │    │            │
       └─────┬────┘    └────────────┘
             │
             ▼
       contextjamming.com
             │
             ▼
       ┌──────────────┐
       │   Git push   │         ← audit trail
       └──────────────┘
Assembled on Mac in Terminal · Filed from Franklin, MAContext Jamming · ACRA Insight LLC · MIT License · FounderFile.ai · RelationalIntelligence.xyz · Commission a Dispatch →