The other side of your term sheet reads faster than you do

On 25 March 2026, Harvey announced a $200 million round at an $11 billion valuation. GIC and Sequoia co-led it. Andreessen Horowitz, Coatue, Kleiner Perkins and Elad Gil joined.

Harvey sells AI agents to law firms. Bloomberg, CNBC and Forbes all covered the round, so the numbers are solid.

Here is the part founders should sit with. The buyers are the firms sitting across from you in a financing.

$11B

Harvey valuation

March 2026 round

$200M

Raised in the round

Co-led by GIC and Sequoia

Equivalent tools most founders use

My observation from the deals I sit in on, not a survey

Sources: CNBC, Bloomberg and Harvey's own announcement, 25 March 2026. The third figure is my read, clearly labelled as such.

The asymmetry was always there

A partner at a fund has seen four hundred term sheets. Their counsel has seen four thousand.

You have seen one, and it is yours.

That gap is not new. Founders have always negotiated against people with more reps. What changed is the cost of closing it on their side, not yours.

A firm with agent tooling reads your documents, compares them against precedent, and flags what is unusual. That used to take an associate a day.

Nothing similar arrived for the founder. Not because it cannot exist. Because nobody sells to a market of one deal.

What actually costs founders money

I sit in on financings. The clauses that hurt are rarely the ones founders ask about.

Everybody asks about valuation. Almost nobody asks about the things below.

Where the value leaks

Clauses founders ask about least, ranked by what they cost

Liquidation preference structure Highest

Anti-dilution mechanism High

Board and protective provisions High

Pro-rata and pay-to-play Medium

Option pool timing Medium

Ranking reflects what I have watched go wrong in deals, not a dataset. Ordering is a judgement call, so argue with it.

A valuation you negotiated up by fifteen percent can be erased by a preference structure you did not read closely. That trade happens quietly and it happens often.

This does not replace your lawyer

Say the obvious thing first. You need a real lawyer for a financing. Nothing below changes that.

What the prompt does is different. It turns a confused hour with your counsel into a focused twenty minutes.

You arrive with questions instead of arriving with a PDF.

Same lawyer, different meeting

Without preparation

Read this and tell me if it is fine

Your counsel explains standard terms from scratch
You pay for the education, not the judgement
You accept the summary because you cannot check it

With a question list

These six things look unusual to me

The hour goes on the two clauses that matter
You can tell when an answer is hand-waving
You know which asks are worth spending goodwill on

The prompt

Paste the term sheet in. Get a question list out.

You are helping a first-time founder prepare to talk to their own lawyer
about a term sheet. You are not giving legal advice and you must say so.

Here is the term sheet:
[PASTE THE FULL TEXT]

Produce four sections.

SECTION 1: PLAIN ENGLISH
For each clause, one sentence explaining what it does in ordinary words.
No legal vocabulary. If a clause has no practical effect in most outcomes,
say "no practical effect in most outcomes" and move on.

SECTION 2: THE MONEY QUESTION
For each clause, answer one question: in which scenario does this cost the
founders money or control? Give the scenario concretely. If it never does,
say so.

SECTION 3: UNUSUAL OR MISSING
List anything that reads as off-market for this stage and geography, and
anything a term sheet at this stage usually contains that is absent here.
Mark each as UNUSUAL or MISSING. State your uncertainty honestly.

SECTION 4: QUESTIONS FOR MY LAWYER
Write 8 questions, ranked by how much money rides on the answer. Each
question must be specific enough that a lawyer can answer it in two
sentences. No general questions.

Do not tell me whether to sign. Do not reassure me.

Section three is where the value sits. A model that has read a lot of financing documents is good at noticing absence. Absence is exactly what a first-time founder cannot see.

The honest limits

The model will get things wrong. It does not know your cap table, your jurisdiction, or what you agreed verbally two weeks ago.

It will occasionally call something off-market that is completely standard where you are raising. Treat every flag as a question, never as a finding.

Used that way it is hard to lose. The worst case is you ask your lawyer a question that turns out to be nothing.

The prompt is on the resources page, free and ungated, along with everything else I give away.