Haga

Data room

External investor sharing

Company Overview

Bridge raise deck

Ready

Temporary bridge track — first check / residency / office. Preserves later $1.0M institutional raise.

Track: temporary soft-raise bridge (micro-check, residency, first institutional believer).
Not the institutional $1.0M / 12 mo deck — keep that for angels/micro-VCs after commercial proof.
Portable pack: soft raise is open, but external dataroom invites stay locked until incorporation — use this MDX + Lab/methodology links for intros (e.g. warm → Brickyard when visa gate is real).

Always state both layers: closing a program-sized tranche now; round target remains $1.0M.

1. Title — bridge objective

Haga
Independent verification for physical AI

Bridge objective: Turn working verification evidence into live customer proof.

Wedge: Fast, private, sim-first physics verification for robot policies and generative world-model outputs — not a public leaderboard, not a sim platform, not a training loop.

Tagline: Stress-test. Break. Report.
Bridge prep · July 2026

2. Problem + narrow wedge

  • Simulated manipulation success does not survive deployment. Stanford HAI 2026 AI Index: 89.4% sim (RLBench) vs 12% on real household tasks — different evaluation contexts, not one controlled system degrading 89→12. Use as urgency signal, not a single-run story.
  • Labs self-report sim scores; synthetic worlds rarely get independent physics checks.
  • Institutional real-robot eval (RoboArena / Robocurve-class) exists at the top — small teams still lack a fast, private, sim-first pre-check.

Wedge (say out loud): sim-first private stress before expensive hardware evaluation — complementary, not a substitute.

3. What is already built

Layer Shipped
Pillar 1 Lift / Stack / PickPlaceCan / Door stress tiers (n=50/condition, Wilson CIs)
Pillar 2 Checker calibration + CogVideoX discovery n=6 (post-hoc) + held-out v1 n=9 (flag 1.000, Wilson [0.701, 1.000], all static_hover)
Trust Public methodology + Lab + article + /methodology
Product Intake → scope → branded partner report workflow + 7 Priority A scopes Ready

This bridge funds commercial proof, not an idea.

Charts: Evidence · Methods: Methodology report

4. One visual proof slide

Primary (equal weight):

Task Nominal → severe Story
Lift 1.00 → 0.26 Grasp slip under stress
Stack 0.96 → 0.20 Grasp fail under stress
PickPlaceCan 1.00 → 0.24 Grasp slip under stress

Boundaries (smaller box — not equal headlines):

  • Door: 0.60 → 0.52 (weak curve; CIs overlap) — show honesty, not hype
  • CogVideoX discovery: n=6, seeds 0–1, static_hover post-hoc — do not cite as held-out
  • CogVideoX held-out v1: n=9, seeds 2–4, flag 1.000, Wilson [0.701, 1.000], all static_hover; real control 0.000
  • Pillar 1 today = scripted OSC baselines in sim — not external learned policies, not RoboArena

5. Commercial readiness vs commercial proof

Ready (shipped) Still to prove (bridge buys this)
Public funnel + methodology Live partner reports delivered
7 Priority A scopes drafted Willingness to pay / LOI / pricing signal
Intake SOP + branded reports External artifact ingestion (customer policy / WM)
Reproducible harness + dry-runs Second-model / external held-out beyond CogVideoX v1

Demand today (honest): Priority A all Introduced; artifact-first wave Introduced; demand-counting conversations = 0. Do not invent LOIs. Progress is tracked on the pipeline.

Outbound bridge re-engage: gated on evidence triggers (external artifact, substantive dialogue, multi-model held-out, grant award, or soft commit) — details shared in direct program conversations, not in this room. Inbound from submitted programs: always reply.

6. Bridge milestones (8–16 weeks)

  1. Delaware C-Corp (Stripe Atlas) + confirmatory IP into the company
  2. Deliver 2–3 live design-partner evaluations (not dry-runs)
  3. Obtain ≥1 paid pilot, LOI, or documented pricing signal
  4. Validate ≥1 partner-supplied or external learned artifact
  5. Held-out Pillar 2 expansion beyond CogVideoX held-out v1 (second model and/or external artifacts)
  6. Continuous-scoring design doc or customer-backed spec (stretch)

Horizon is weeks, not the 12-month institutional operating plan.

7. Bridge budget posture

Do not reuse the $1.0M percentage allocation for micro-checks.

Frame spend as milestone cost + short runway:

Bucket Intent (bridge)
Entity / legal Atlas + counsel for SAFE / confirmatory IP
Founder runway Months of full-time focus through live reports
Compute Targeted Pillar 2 batches — not $250k institutional GPU plan
GTM Narrow travel / partner calls only after reply heat
Buffer Diligence spikes

Exact dollars = program-specific (see ask variants below). Monthly cash need stated verbally to match check size.

8. Why this unlocks the later $1.0M

Bridge converts technical proof into customer and pricing proof. The $1.0M pre-seed then funds repeatable delivery, first ML/eval hire, continuous-scoring productization, and expansion beyond founder-led services.

Institutional milestones remain locked: 2 partner reports ≤4 mo · Pillar 2 n≥30 ≤6 mo · continuous-scoring design doc ≤6 mo · first hire/co-founder offer ≤9 mo — Milestone model.

9. Team + constraints

Sole founder: Mushood Hanif — shipped dual-pillar harness, public Lab, intake→report product.

  • Cap table intent at incorporation: 90% founder / 10% option pool — Founder equity & IP (Gate A signed; entity still pending)
  • Hire / co-founder search on written scorecard — join via new grants, not placeholder roster
  • Location / work auth (one line): Building from Pakistan; willing to relocate for the right program; US work authorization is a planned step after entity + capital — Delaware incorporation ≠ visa
  • Close path: soft commit → Atlas Delaware → SAFE / wire into company bank — Formation

10. Ask — bridge (structure only)

Rule: state both layers — closing a program-sized tranche now; round target remains $1.0M.

Generic bridge framing (angels / micro-VCs / first believer):

  • Instrument: post-money SAFE into Delaware C-Corp after Atlas
  • Size: match investor (typically $25k–$250k tranche; residencies vary)
  • Use: entity + short runway + compute for live partner reports + held-out Pillar 2
  • Institutional target unchanged: $1.0M / ~12 months

Program-specific variants (Unshackled, residencies, regional accelerators, Brickyard-class leads) are not published in this room — they signal application status and weaken parallel conversations. Share tailored slides in direct meetings when the program or intro is real. Rationale: Disclosure policy.

Appendix — claim boundaries (say out loud)

  • No live customers, completed customer evals, LOIs, revenue, or validated pricing yet

  • Pillar 1 = scripted OSC baselines in sim — not external learned policies; not a RoboArena substitute

  • CogVideoX discovery = n=6 post-hoc; held-out v1 = n=9 separate (static_hover); not multi-model yet

  • Gate A signed ≠ Delaware company owns perfected IP title yet

  • Soft raise materials ready ≠ outbound re-engage unlocked (evidence-trigger gate) · dataroom auth still locked pre-incorporation

  • Delaware incorporation ≠ US residency / work authorization

  • Competition claim: fast, private, sim-first dual-artifact eval — not “nobody checks physics”

Institutional deck (later): Pitch deck · Disclosure: Disclosure policy