AthenisHQ An AI-native studio solving accuracy
3 systems live today. 3 more building in public.

Software engineered for accuracy: from AI behavior to daily workflows.

AthenisHQ builds intelligent tools designed for high-value tasks, from AI behavioral evaluation and financial analysis to real-world visual inspection.

See the live products ↓
6Products
3Live in production
3In development
LAST VERIFIED —
Products
AletheiaAI Infrastructure & Safety
Operational

Behavioral observability for production AI — 9 signatures, empirical failure rates instead of vibes.

2,780Incidents validated
9Signatures tracked
→
Aletheia MCPAI Infrastructure & Safety
Open source

Runtime filter that blocks scope-creep and prompt-injection tool calls before an AI agent runs them — deterministic, no LLM in the hot path.

~25 µsFilter overhead
155Tests passing
→
OwlScoreFintech & Capital Markets
Operational

AI stock scores hash-chained to a public ledger the moment they're published — 59 predictions live.

43.6%Backtest accuracy
59Live predictions
→
OwlNoteProductivity Tools
In development

Handwritten notes for iPad and iPhone — a freeform Pencil canvas with an AI assistant that reads your handwriting.

SkiaRender engine
2Devices supported
→
CAL-CProductivity Tools
In development

An AI smart calculator with 18+ specialized tools and a reactive tape history.

18+Specialized tools
10 dec.Precision
→
TireScanMobility & Computer Vision
In development

Snap a photo of your tires, get an instant AI read on tread depth and wear.

iOSPlatform
Capabilities

What each product actually does.

Aletheia
  • Detects 9 distinct behavioral signatures in LLM output — confidence without grounding, scope creep, credibility exploitation, and more
  • Every signature validated against 2,780 real-world incidents from AIID, AVID, and the MIT AI Risk Repository
  • Public benchmark scores and a live playground — nothing is a private claim
Aletheia MCP
  • Blocks known scope-creep and prompt-injection tool calls before an agent runs them — destructive shell/SQL, credential reads, SSRF, out-of-mandate writes
  • Deterministic, ~25 µs per call, no LLM in the hot path; standalone guard tools or a fail-closed proxy over any MCP server
  • Apache-2.0, 155 tests, on npm and the official MCP registry — a first-line filter, not a sandbox
OwlScore
  • Scores the S&P 100 nightly and hash-chains every score to a public ledger before the outcome is known
  • 43.6% historical backtest accuracy, tracked and published, not cherry-picked
  • 59 live forward predictions currently open — screener, watchlist, and full track record are public
OwlNote
  • Freeform Apple Pencil canvas built on a Skia render engine, tuned for iPad and iPhone
  • An AI assistant that reads handwritten strokes directly — no forced transcription step
  • Folders, direct drag-to-move, and per-note typography themes
CAL-C
  • 18+ specialized calculator tools beyond arithmetic — unit conversion, finance, statistics
  • Math held to 10 decimal places on every operation, not floating-point approximation
  • A reactive tape history you can edit and recompute from any prior step
TireScan
  • Reads tread depth and wear pattern from a single photo, no special equipment
  • Flags uneven wear that usually indicates an alignment or suspension issue
Recent activity

Aletheia MCP — published to npm and the official MCP registry; runtime tool-call filter now installable with a single command.

OwlScore — 59th live prediction added to the public ledger. Backtest accuracy holds at 43.6% against S&P 100.

Aletheia — behavioral taxonomy validated against 2,780 real-world incidents across AIID, AVID, and the MIT AI Risk Repository.

OwlNote — Pencil canvas and folder system reach feature parity across iPad and iPhone.

CAL-C — 18th specialized calculator engine shipped to the internal build.

Studio

Six products deep, three shipped and live, before the first outside hire.

Team

AthenisHQ is led by a founding team of engineering leaders with over a decade building AI-powered platforms and infrastructure at internet scale — leading teams that ship products used by hundreds of millions of people, and building the deployment and reliability systems that keep them running.

Standard

Production-grade, at scale, no shortcuts — the bar every product is held to. Three are already live proving it end-to-end; the rest are close behind, and the studio is now building the team to take all six further.

None of the six compete on being first — they compete on being verifiably right: a public hash-chained ledger instead of a claimed track record, empirical failure rates instead of vibes-based safety claims.

Monetization
  • Aletheia — source-available, paid commercial licensing for hosted use
  • Aletheia MCP — Apache-2.0, free and open source
  • OwlScore — free, full S&P 500 coverage planned as a premium tier
  • OwlNote & CAL-C — free-to-start, Pro subscription unlocks sync and AI

Get in touch

Questions about a product, a partnership, press, or something else entirely — reach out directly.