Active Projects

Projects with defined outcomes requiring multiple steps. Reviewed during cowork sessions.

Hierarchy: Vision projects define where we're going. Ops projects keep the business running and fund the vision. Vision always has priority.

The Triangle:

          HIHQ (products)
         /              \
        /  intelligence   \
       /   & credibility   \
      /                      \
Angel ——— lead gen ——— Worca
(deal flow)   & talent   (cashflow)

Structure:

• HIHQ = data & AI infra research lab. Long-term, no VC, no exit. Builds eval/RLHF products for Big 6.
• Worca = cashflow engine. Funds HIHQ. Also a data source (HR vertical). Proving ground for AI ops.
• Angel = intelligence & distribution arm. Deal flow + domain access + lead gen + credibility. 1000x returns are real and standalone.

Vision — HIHQ

Domain-specific AI evaluation & data research lab. Self-funded by Worca cashflow. No VC, no exit plans. Build real IP, publish, become the authority.

HIHQ North Star

• Owner: Steven
• What HIHQ Is: A data & AI infrastructure research lab that captures scarce human expert judgment in hard-to-access domains and packages it as evals, benchmarks, and RLHF datasets for the Big 6 AI labs (Google DeepMind, Anthropic, OpenAI, xAI, Meta, Amazon).
• Founder-Market Fit: Taiwanese roots, Science Park network, deep semiconductor industry access. Very few people sit at the intersection of "AI eval methodology" + "access to domain experts in fields where ground truth is scarce."
• Edge: Domain-specific evals in verticals where expert judgment is rare and expensive. Generic eval shops (Scale AI, LMSYS) can't compete on domains they can't access.
• Verticals:
  1. Semiconductor — first vertical. PhD-level semiconductor experts with deep AI passion. IC design, process engineering, fab operations. Steven's Taiwan + Science Park network is the moat.
  2. HR & Talent — powered by Worca. Recruiting judgment, talent-client matching, cultural fit assessment.
  3. Future verticals — any domain where human expert judgment is scarce and valuable (legal, medical, finance). Same playbook, new domains.
• Sequencing:
  1. Evals & Benchmarks — publish domain-specific benchmarks (e.g., "Can frontier models assess semiconductor talent as well as a TSMC veteran?"). Gets HIHQ in the door, builds credibility with lab eval teams.
  2. RLHF Datasets — preference pairs from real expert decisions. Bigger contracts, longer relationships.
  3. Multi-vertical expansion — apply the capture playbook to new domains.
• Publishing Strategy: Publish benchmarks and research openly. Lab ML teams don't take sales calls — they read papers. A published benchmark circulates inside every lab's eval team. The sales call books itself.
• Funded By: Worca cashflow. No dilution, no investor pressure, no exit timeline.
• Updated: 2026-02-09

H1: Semiconductor Eval Data

• Vertical: Semiconductor
• Outcome: First domain-specific eval dataset using PhD-level semiconductor experts
• Owner: Steven
• Status: 🟡 In Progress — Semiconductor Experts ops project (below) is simultaneously building this dataset
• Why First: Founder-market fit. Taiwan + Science Park network = access nobody else has. Semiconductor expertise is scarce, expensive, and exactly what labs need for domain-specific eval.
• Next Action: Define what semiconductor eval benchmarks look like for the labs
• Updated: 2026-02-09

H2: HR & Talent Eval Data

• Vertical: HR & Talent
• Outcome: Packaged eval/RLHF datasets from recruiting judgment, ready to sell to Big 6
• Owner: Steven
• Status: 🟡 Design Phase
• Data Source — Worca: Worca's daily recruiting operations are HIHQ's data pipeline for this vertical. Every client call, every talent assessment, every match decision Worca's team makes is raw material for eval/RLHF datasets. Worca doesn't need to change what it does — HIHQ captures and packages the judgment that's already happening. As Worca acquires more agencies (W2), the data volume scales automatically.
• Data Assets:
  1. LinkedIn + public info analysis and annotation → AI-generated reports with human scoring
  2. Meeting recordings — discovery and closing calls
  3. Q&A during discovery and closing calls
  4. Client feedback on talent quality
  5. Stated decision-making speed vs actual
  6. Company signals
  7. Personal goals and motivations
• Approach: Evaluation first (build scoring rubrics + benchmarks), then RLHF (preference pairs from real recruiter judgments)
• Next Action: Define what the data product looks like — what format do labs buy eval/RLHF data in?
• Updated: 2026-02-09

H3: Published Benchmarks & Research

• Outcome: HIHQ publishes domain-specific AI evaluation benchmarks that lab eval teams cite and use
• Owner: Steven
• Status: ⏳ Blocked — needs H1 or H2 to produce initial datasets first
• Depends On: H1 (semiconductor data) or H2 (HR data) — need real data before publishing
• Next Action: None yet — focus on H1 and H2
• Updated: 2026-02-09

Vision — Worca

Cashflow engine. Funds HIHQ. Proving ground for AI ops.

Worca Strategic North Star

• Owner: Steven
• Goal: Agent & AI Arbitrage. Acquire agencies with inefficient ops, use AI to reduce costs to near-zero. Buy at staffing multiples, operate at near-zero marginal cost.
• Role in Triangle: Worca makes money. That cashflow funds HIHQ. Worca's recruiting operations also generate HR & talent eval data as a byproduct (feeds HIHQ H2), but that's HIHQ's concern — Worca's job is cashflow and growth.
• Flywheel: AI cuts ops costs → higher margins → acquire more agencies → more cashflow → fund HIHQ → repeat
• Updated: 2026-02-09

W1: AI Ops Infrastructure

• Goal: Agent & AI Arbitrage
• Outcome: Prove AI can run agency operations at near-zero cost — first on Worca, then on acquired agencies
• Owner: Steven
• Status: 🟡 In Progress — Alfred + agents operational, not yet applied to full ops workflow
• What "Done" Looks Like: Worca's internal ops (triage, sourcing, matching, scheduling) run primarily through AI agents with human oversight
• Next Action: Identify which ops workflows have the highest cost and are most automatable
• Updated: 2026-02-09

W2: Agency Roll-Up

• Goal: AI Arbitrage
• Outcome: Acquire first agency using the proven AI ops playbook
• Owner: Steven
• Status: ⏳ Blocked — requires W1 proven first
• Depends On: W1 (proven ops cost reduction)
• Next Action: None yet — focus on W1
• Updated: 2026-02-09

Vision — Angel

Intelligence & distribution arm. Every investment compounds across all three entities.

Angel North Star

• Owner: Steven
• Primary Goal: 1000x returns. Back founders who know something the market doesn't. This is real and standalone.
• Strategic Value to the Triangle:
  1. Lead gen for Worca — portfolio companies need to hire. Invest → founder trusts you → Worca places talent. Built-in sales channel at zero marginal cost.
  2. Intelligence for HIHQ — deal flow is one of the highest-signal info streams in tech. You see what AI companies are building, what data they're struggling to source, what capabilities are missing. That's product roadmap intel for HIHQ.
  3. Domain access for HIHQ — every portfolio company opens a door into their domain's expert network. Autonomous Investing is the proof: angel investment + Worca revenue (Semiconductor Experts) + HIHQ's first semiconductor dataset. Triple play. Future investments in healthcare AI, legal AI, etc. open new HIHQ verticals.
  4. Credibility with Big 6 — "We invest in AI infrastructure companies and operate a domain-specific data lab" lands differently than "we're a staffing agency with data." The portfolio signals you understand their world.
  5. Talent recycling — portfolio companies that fail release talented people → Worca places them. Portfolio companies that succeed need more people → Worca places them. Either outcome feeds Worca.
  6. Diligence as data — every founder screened, every technical deep dive, every reference call is judgment data about evaluating technical talent and companies. Exactly the kind of eval signal HIHQ packages.
• Investment Focus: AI Data Infrastructure, Post-Transformer Architectures, Computer Vision + Camera Tech. See agent-angel/context/INVESTMENT_THESIS.md for full criteria.
• Check Size: $10K-$50K per deal, Seed/Pre-Seed. Self-funded.
• Updated: 2026-02-09

A1: Build Portfolio with Strategic Overlap

• Outcome: 10-15 investments where each company creates value across the triangle (returns + Worca lead gen + HIHQ domain access)
• Owner: Steven
• Status: 🟡 In Progress — 5 portfolio companies, 6 in active diligence
• Current Portfolio: Ambiguous, Autonomous Investing, Bluebird, Metropolis, Yaw
• Triple Play Example: Autonomous Investing = angel investment + Worca client (Semiconductor Experts) + HIHQ semiconductor data source
• Next Action: Evaluate active pipeline (Boxsy, Gro, HumaneBench, Hydrolyca, Saola, StealthNet) through triangle lens — which ones create multi-entity value?
• Updated: 2026-02-09

A2: Founders Watchlist

• Outcome: Be first check when Tier 1 founders start their next company
• Owner: Steven
• Status: 🟢 Tracking
• Watching: Daniel Smilkov (ex-Lilac/Databricks), Kojin Oshiba (ex-Robust Intelligence/Cisco), Nikhil Thorat (ex-Lilac/Databricks), Omer Ayubi (OpenAI)
• Next Action: Monitor for departure signals. Move fast when they raise.
• Updated: 2026-02-09

Ops

Day-to-day projects that keep Worca running and generating revenue. These fund the vision.

Semiconductor Experts (Autonomous Investing / Hengxin)

• Outcome: Source and place PhD-level semiconductor experts for Hengxin/Autonomous Investing — domain experts passionate about AI, not gig workers
• Feeds Vision: Triple play — Angel (Autonomous Investing is portfolio company) + Worca (client revenue) + HIHQ H1 (first semiconductor eval dataset). The single most strategically important ops project.
• Owner: Steven + Alec (Taiwan operation)
• Sourcing: Lydia + Linda (interviewing candidates)
• Status: 🔴 Critical — make or break for Worca & HIHQ 2026
• Next Action: Brief team that this is #1 ops priority above all else
• Signals: [TBD]
• Updated: 2026-02-09

Bees360 Growth (Major PH Client)

• Outcome: Grow Bees360 as major client in Philippines
• Owner: Steven + Andy Liu
• Status: 🟡 In Progress — booking flights
• Key Milestone: PH trip Mar 2-4 to visit Joel (local partner)
• Plan: Andy flies to LA, then fly together LAX → Manila. Depart Sat Feb 28 evening, arrive Mon Mar 2 morning. Return Wed Mar 4.
• Next Action: Confirm flight with Andy, book tickets
• Open: Richard's availability still unconfirmed
• Updated: 2026-02-11

Completed

• Helena March Launch — Resolved (2026-02-11)
• Remote Payment Resolution — Resolved (2026-02-11)
• Jaspher/Rune Modeling Project — Resolved (2026-02-11)
• Benji Onboarding — Resolved (2026-02-11)
• Cymate Exit — Resolved (2026-02-11)

Someday/Maybe

Parked projects — review monthly.

Insurance Coverage Gaps (iHerb COI)

• Outcome: Close two coverage gaps exposed by iHerb auditor requests
• Gap 1 — Professional Liability / E&O: Not covered under current Hiscox BOP (explicitly excluded). Sharon Drabeck asked if PL is in GL or separate — it's neither. Need standalone E&O policy. Get quote from Hiscox (844-357-0403 or hiscox.com/manage-your-policy). Told Sharon we're working on it (Feb 11).
• Gap 2 — Workers' Compensation: Not covered under current Hiscox BOP (explicitly excluded). Chels/Sharon originally requested this COI (Jan 29). Need to determine if required by state law, then get quotes (Hiscox, State Fund, or PEO like Deel/Remote).
• Current policy: Hiscox BOP P104586161, GL $1M/$2M aggregate + Property $100K. Effective 03/11/2026. Premium $2,666/yr.
• INSURANCE_SUMMARY.md was incorrect (claimed E&O was covered) — corrected Feb 11.
• Source: Email from Sharon (Jan 29) — "Request COI"
• Review: Feb 15, 2026
• Updated: 2026-02-11

iHerb Ariba Vendor Registration

• Outcome: Register Worca on iHerb's Ariba Network for vendor payments
• Why Parked: Administrative task, not urgent — waiting for Tiara to send registration link
• Action Needed: Provide admin name + email to Tiara Brooks (tiara.brooks@iherb.com)
• Source: Email from Tiara Brooks (Feb 2) — "Worca Jan payment"
• Review: Feb 15, 2026