W/ AXIS
YC FALL 2026WORKING MVP

PERSISTENT JUDGMENT MEMORY · FRONTIER INVESTING

An AI research team that learns how investors judge.

W Axis monitors frontier signals, presents the evidence and strongest counter-case, and learns from every accept, reject, edit, action and outcome.

Run the judgment loop ↘Watch 30-second demo ▶
01 Solo founder02 Working interactive MVP03 Seeded public demo
DECISION OBJECT / W–017SPACE × ENERGY
STRUCTURAL SCORE91/100

Surface power is becoming the operating system of the lunar economy

Power conditioning, thermal rejection and autonomous distribution may become compulsory infrastructure for every persistent lunar asset.

01 / EVIDENCEA real bill is forming below the visible product.
02 / DISCONFIRMProgram timing and customer concentration remain unresolved.
03 / CONTROLPower electronics + autonomous distribution.
04 / NEXT ACTIONVerify awarded hardware before narrative exposure.

01 / THE PRODUCT

AI has enough information. It lacks a record of why a person decided.

Chat memory preserves conversation. W Axis stores structured decision traces: signal, rule, evidence threshold, contrary case, decision, action and outcome.

GENERAL AI RESEARCH
  • Retrieves and summarizes
  • Starts again from the next prompt
  • Optimizes for a plausible answer
W AXIS
  • Remembers accept, reject, edit and outcome
  • Changes future search and ranking
  • Seeks disconfirmation before action

02 / THE LEARNING LOOP

One repeatable path from weak signal to approved action.

01

Observe

Monitor weak signals before a stable category forms.

02

Test

Run each signal against the user’s thesis and evidence threshold.

03

Disconfirm

Surface the strongest reason the thesis may be wrong.

04

Decide

Capture advance, investigate, reject, edit and approved action.

05

Learn

Use the decision and outcome to change every future ranking.

03 / PRODUCT FILM

See the judgment memory update in thirty seconds.

This short cut uses Amy’s authorized voice clone and a visibly seeded dataset. The interface interaction is the proof; the example count is not traction.

04 / LIVE WORKSPACE · SEEDED DEMO

Teach the system one decision.

Select a signal, inspect its evidence and counter-case, then decide. The interface updates immediately to show what W Axis would remember.

W–017SPACE × ENERGYPRE-CONSENSUS

Surface power is becoming the operating system of the lunar economy

Power conditioning, thermal rejection and autonomous distribution may become compulsory infrastructure for every persistent lunar asset.

STRUCTURAL SCORE91/100
Real bill HighControl layer VisibleTiming Early
01 / EVIDENCE CHAIN
  • Mission architectures are converging on persistent surface operations.
  • Power availability constrains communications, compute and payload duty cycles at once.
  • The supplier map is fragmented and several control-layer vendors are still misclassified.
02 / STRONGEST COUNTER-CASE

Program timing can slip for years; concentrated customers and qualification cycles may erase an early-mover advantage.

03 / CONTROL LAYER

Power-management electronics + autonomous microgrid software

TEACH YOUR JUDGMENTWhat should W Axis do with this signal?
NEXT REAL-WORLD ACTION

Map 12 suppliers, request two founder calls and watch awarded hardware—not announced concepts.

05 / STARTING WEDGE

Small frontier-tech funds and family offices without a large analyst team.

They already have access to information. Their edge lives in tacit judgment: which evidence they trust, what makes an opportunity too crowded, and when a signal deserves a founder call, investigation or investment review.

FIRST WORKSPACEFrontier-investment research
PRIMARY OBJECTDecision + outcome history
LONG-TERM LAYERJudgment memory for AI agents
Amy Li, founder of W AxisAMY LI · FOUNDER & CEO

06 / FOUNDER

Built from a real frontier-investing workflow—not a whiteboard category.

Amy previously worked across Bitcoin mining, proof-of-work infrastructure, crypto ecosystems and frontier-technology investing. She built the underlying system to decide before markets had stable categories for what was changing.

“The valuable asset was not another report. It was the history of why I trusted one signal, rejected another, and decided to act.”
Solo founderFull-time since Dec 2025NYC / Beijing → San Francisco

07 / PRODUCT STATUS

Working now. Honest about what comes next.

WORKING MVP
  • Ranked frontier-signal queue
  • Evidence and disconfirming case
  • Advance / Investigate / Reject memory
  • Judgment-shaped investigation route
NEXT PRODUCTION MILESTONES
  • Live source ingestion
  • Durable multi-user memory
  • Authentication and permissions
  • Decision-and-outcome evaluation
DEMO DISCLOSURE

Signals and the initial memory count are seeded. The decision interaction is real. Seeded values are not users, revenue or traction.