Outside the firm

Enkodia AI

AI startup pitching an "Unlimited Language Model" for legal and accounting; deck shared with Brandon Drespling by a contact on or before 2026-09-01
Where things stand
  • Pre-UI, pre-MVP, pre-revenue startup raising $1–5M to build a UI and MVP and hire engineers around inventor Trevor Chandler; first market legal and accounting, then finance, academia, medicine.
  • Team per the deck: Luke Fisher (co-founder, Stanford student, deck contact), Trevor Chandler (inventor, 30+ patents), Adam Pener (0-to-1 operator, prior collaborator with Chandler), Randy Roberts (Harvard-trained Denver attorney and investor).
  • JT's read as of 2026-09-03: real problem, oversold product, no third-party test, not underwritable from the PDF.
  • Deck reached Brandon through a contact; who that is and what they want from Brandon is not on file.
Relationship to JT
JT reviewed the deck at Brandon's request; no direct contact with Enkodia
Channels
via Brandon
Notion tasks
none
Related files
Enkodia deck review · Enkodia Pitch Deck Draft (2).pdf (not read; linked) · Brandon-Note v4
Last meeting
none
Last updated
2026-09-03

Focused on

  • Raising $1–5M (range still marked "[consult Trevor on what he would need]") to build a UI, MVP, and feature work; usage-based pricing first, then about $1,500/user/year subscriptions. Enkodia review
  • Claims: not a wrapper, unlimited context, 0% hallucination, flat token cost, no KV-cache growth, runs locally on any GPU, "Panel of Experts". Enkodia review

Asking me for

Nothing yet.

Concerns and frustrations

  • Their traction slide admits "significant expressed interest" but that skepticism blocked anything being put in place; third-party benchmarks refused as proprietary risk. Enkodia review

Doable vs. not doable

  • JT's read: the pain is real; the claimed solution is oversold, untested, and priced against a fictional competitor. To be more than a deck they would need a reproducible demo on hardware and documents of the reviewer's choosing, honest pricing against Sonnet 1M and self-hosted open weights, a narrow first wedge, a named design partner, and a team that can sell into firms. Enkodia review §9
  • JT's read on the 0% hallucination claim: only true if the product is rails plus a file library, a program rather than a new model, which is real work but not unique and only sells to shops that do not know they can build it. Brandon-Note v4

How to work with them

  • If anyone diligences them, use the ten questions in the review §10: demo on our hardware with the network off, operational definition of hallucination, reconcile the 2,200× vs 5,500× cost slides, cost at 32K context vs Sonnet and self-hosted Qwen/DeepSeek, architecture diagram, which patents cover this system, why legal first with no Westlaw/Lexis license, what Trevor needs in dollars and months, who already said no. Enkodia review

Crossover

  • touches Brandon Drespling: Brandon owns the relationship with whoever sent the deck and is deciding what to say back; JT's guidance to him is in the note draft. Brandon-Note v4

Decisions log

Nothing yet.

Meeting log

Nothing yet.

History

Nothing yet.