What’s new

Settings, Firm has a new card: each night the room can send us totals of which pages and features were used, how often and by how many people, so we can decide what to build and fix. It never includes a deal, a company, a contact, a name, a figure or anything you typed, and it never says who on your team did what. The card shows the exact totals that would be sent next, and turns sharing off.

The staff now have the titles a real venture firm uses: the Investment Analyst, the Screening Associate, the Devil's Advocate, the Valuation Analyst, General Counsel, the Portfolio Manager and the rest. The Merit read on a deal comes from the Independent Reviewer, and the Performance Analyst scores its calls. The firm-wide chat is now Ask the firm, and it answers how-to questions from the Help page.

Every room now has a Help page: articles for each page, a question box that answers only from them, and a way to reach a person. It cannot see your deals.

The Trust page now separates three things: what automated tests keep true on every release, what has been measured on real AI output, and what is not proven yet. It no longer reads as proof that an analysis is right.

Every deal now shows one call, Invest, Watch, Pass or Not enough info, with one reason and one next step, in the same words in the Deals list and at the top of the deal. The working behind it has moved to the Analysis tab. A deal is only called out of your mandate when it contradicts something you set, and that reason is stated plainly.

When a founder applies, the first analysis starts straight away on every plan, and the deal shows Analysing until the call is in. Each plan has a daily limit on these analyses, and you can set a lower one. Your off switch still applies.

The owner can upload a logo under Firm settings, and the sidebar shows it in place of the firm's name. The file is adjusted first: a plain background is removed, the margins are trimmed, and a logo whose colours would not read on your sidebar is shown in white instead. The settings page previews it on the sidebar and says which was done. Change the firm's colour and the logo is fitted again, with nothing to re-upload. Without a logo, the sidebar shows your name as before.

The team's morning email now comes from the daily meeting, sent when it closes: the Chief's one line, the state of the firm in two or three sentences, the numbers, the decisions that are yours alone, the deal of the day, and one line per area. What the staff is handling or has parked is a count rather than a list, and the full list stays on the Needs attention page, linked from the email. If the meeting does not convene, the email still arrives and says so.

The convergence meeting the staff hold each cycle now follows the same agenda every day: new since yesterday, the round-table where every seat shares what it found, the decisions by area, the one contested deal the room argues to a conclusion, the follow-up reads the staff picked up, and a closing retrospective. The retrospective is grounded in the cycle's own figures — what the firm resolved and argued to a conclusion (what went well), where the queue of open work grew or deals sat at the decision point (what didn't), and one concrete thing to improve next cycle. It stays honest on a quiet day rather than manufacturing a win. The full agenda, each item showing what filled it, sits on the meeting page behind the morning briefing.

gBizINFO, EDINET and J-PlatPat are now default sources the AI may draw from — Japan's government record of subsidies and public contracts, the securities filings of listed Japanese companies, and the Patent Office's patents and trademarks. When a deal rests on a Japanese grant, a listed customer or partner, or a Japanese patent, the staff can look it up by the company's corporate number, in Japanese, and cite the record; a claim with nothing on file comes back as not found rather than disputed, and none of them is treated as confirming a private round. Switch any of them off in Settings like any other source.

The valuation card now sets three figures side by side, all pre-money so they actually compare: what similar rounds are paying, what the Valuation Analyst independently calculates the company is worth, and what the founder is asking — with the ask read against both. A mark on the Valuation Analyst's price opens the formal methods behind it — the dilution backsolve, comparables, Payne's Scorecard, Berkus, a revenue multiple — each with its own number and shown as cross-checks rather than blended into one, so the figure is one you can put in front of an LP. Where no round is recorded but the deck states a price, the card now shows it, marked deck-stated, instead of reading as if there were no ask at all.

A raise recorded in US dollars for a company based in a market with its own currency could be read in that local currency — a dollar raise on a Japan-based company was priced as yen, pulling the computed valuation far below reality. The raise is now priced in the currency it was actually stated in, so the Valuation Analyst's number is right wherever the company sits.

SEC EDGAR (sec.gov) is now one of the default sources the AI may draw from, alongside the other registries and regulators — the checkable record for a claimed US funding round. When a deal rests on a US round, the Diligence Associate, Valuation Analyst and Technical Advisor can look for the matching Form D by the issuer's legal name, compare its date and amount to the claim, and cite it; a round with no filing comes back as not found rather than disputed, since many SAFEs and small seed rounds are never filed. Switch it off in Settings like any other source.

Your first read on a deal — the house's own invest, watch or pass — now pulls the independent reads in with it. They raise during screening, so they inform the decision to advance rather than arriving after you have already made it. And when the first read is positive — invest, or watch at real conviction — the reads your room has run on their own, so the opposing case and the price are in hand before you advance; on a pass, or a watch you barely hold, nothing runs. Where a read still needs a deliberate start, it is one button press.

A deck that arrived but never appeared in the data room — so the Investment Analyst and the Diligence Associate never read it — now gets filed on its own the next time the room is opened, and the reads that depend on it can run. Any deal whose deck got stuck this way heals itself on the next visit rather than sitting silently unread.

A deal’s live forecast refreshes against the latest reads instead of going stale after the first pass, and it now sits inside the track record where its calls can be scored over time rather than in a panel of its own. The number a partner reads is the current one, and its history is where history belongs.

The relationship graph is now read per deal: it surfaces who to syndicate a round with, which of your contacts opens a warm path to the company, and where an introduction is worth making. A partner working a deal gets the people half of it beside the analysis instead of having to remember who knows whom.

When someone you shared a deck or data room with re-opens it several times in a short window and that deck ties to a live deal, the Syndicate Manager now raises it — a contact who opened a deal's deck four times this week is worth a call while it is warm. It is a sign that interest is heating up, surfaced while it still means something. A note for you, not a move it makes on your behalf.

The IC Secretary used to wait on the track-record page for you to come and look. It now brings a pattern to you when one side of your record is worth facing — fewer than half your invested deals working out, say, or fewer than half your passes proving right — once there are enough settled calls to mean something, and it stays quiet on a healthy record. And when a founder you backed has a fresh win while you have gone quiet with them, its verdict carries a nudge to reconnect — never for a founder you have marked do-not-contact.

A deal’s summary now leads with the three questions a partner is actually deciding — is it worth the money, do these terms make sense, and is it a fit for this firm — each answered by the seat whose job it is, and each sourced. The reasons sit in one place rather than being pieced together from the memo, the forecast and the pricing.

On a deal’s decision view, the valuation seat now answers the two questions a partner is actually deciding: whether the business is worth the money, from its own sourced price band, and whether the proposed terms make sense once an ask is on file. Its verdict sits on the decision screen beside the forecast and the memo, so the pricing call is read in one place rather than pieced together.

Sharing a deck or a data room view-only reads more clearly now: the link expiry says "Expires in 15 days — blank means never" instead of a bare number, the NDA, email and consent gates are grouped under a Protections heading apart from the slideshow display toggle, and the founder-consent line is a quiet note rather than a loud banner. What the link actually allows is unchanged — only how the choices read.

The attention list no longer lets an item drift to the top just for being old. Every open thing is weighed on two separate axes — how much it matters, and how little time is left to act — so a large decision due next week outranks a small one that has only been sitting. A partner opening the list sees what to do now, for a reason the list can name.

When a seat proposes something — an introduction, a letter, a next step — it becomes a first-class action with a life of its own: proposed, then sent, then its outcome. A deal reads “raised, worked, landed” instead of a pile of suggestions nobody closed, so a partner can see what was acted on and what is still owed.

You can now add a teammate as a read-only reviewer who sees the pipeline and the analysis but cannot change anything, so an outside eye or a cautious new hire can look without being handed the keys. And a workspace now requires a reachable owner address, which means the forgot-password path is always there when you need it.

A mutual NDA could get stuck reading "not ready — unmapped field" when the live letter still named the counterparty's notice address by an older field name. That name is now recognised as the counterparty's email, so the letter is ready to send instead of blocked — and a room whose NDA still carries the old field is fixed without anyone having to touch the letter.

A few places that read as broken now read true. The Chief's morning brief no longer leaks internal reference codes into its prose. Two things at risk on one company show as two distinct findings, each naming what it is, instead of identical-looking twins. A key-question card with no upside case on file says so, rather than showing an empty box. And a firm-level finding — firm settings incomplete, a cold-LP nudge — now carries a working action that takes you where you can fix it, instead of a dead "open the deal" with no deal to open.

A position that has exited now values from what actually happened — the proceeds on an acquisition, IPO or write-off, and only the remaining stake after a partial secondary — instead of stacking a stale paper mark on top of the result. This corrects the fund multiple and the win-and-loss record a partner shows investors, so a write-off reads as a loss and not a win.

The weekly founder check-in now asks each founder how the company is really doing, and that reading flows straight into the portfolio view. A partner sees the founder’s own health signal next to the figures already on file, so a company drifting off-track surfaces from what the founder says and not only from the metrics.

When a position has no valuation date on file, the Secondaries Associate no longer invents a timeline from it and no longer prints an impossible figure for how long it has been held or what it is worth. It says there is no mark on file and makes no claim — one less number on the page a partner has to distrust.

Every colleague used to work on one cadence, the nightly round, so a deck that landed at nine in the morning waited until the small hours to be read. The seats that screen a new deal now wake when the deck arrives: the Investment Analyst's read, the Technical Advisor's outside check, the Diligence Associate's trace and the Devil's Advocate's objection are waiting minutes later instead of tomorrow. You choose which seats wake on what, under Settings, and the nightly round stays on regardless.

The nightly round now starts four hours earlier. With the reasoning seats spread across the night and the Chief waiting on all of them, its brief could previously land after a partner had already opened the queue. It finishes well before the morning now.

The Diligence Associate, the Reference Checker, General Counsel and the Market Analyst were entitled, switched on, and had work waiting, and none of them had run once. The nightly round works its seats in order inside one time limit, these four are last, and the time was gone before their turn, while a stopped run left nothing behind to show it. Each seat now gets a fair share of the time remaining, and one that overruns is recorded as timed out while the round carries on.

When you decline with a note to reconnect, the follow-up used to be dated 120 days out whatever the letter said, so a promise to reconnect in Q3 became a task in November. The date is now read from the timing the letter actually states. Where the revisit is a condition rather than a date, the task says so and asks you to set one.

A portfolio company's problem is usually one introduction away from somebody you already know. The Operating Partner reads what a company needs, searches your relationships for the person who can help, and drafts the introduction. You send it or you do not.

The seat that paired investors with deals now works the capital side end to end: interest that was promised and never hardened, allocations that have stalled, and LPs who have gone quiet since the last close. The name changed with the job, and nothing you have configured moves.

Some deals need a language and a context the rest of the staff does not carry. The Japan Desk translates the deal and reads the room around it, per deal rather than nightly, so it is there for the deal that needs it rather than running over everything.

A firm that does not grade its own decisions repeats them. The IC Secretary logs every decision as it is taken and returns to it the moment there is a verdict to compare against, so the record is built while it happens rather than reconstructed a year later. It closes the loop the Independent Reviewer opens.

Liquidity windows open and shut without announcing themselves, and a holding nobody is watching is a holding you sell late. The Secondaries Associate follows what is moving around each position and raises the moment worth a conversation. It never acts on a holding, it tells you.

Every read now carries a control that runs the reader two to five times over the same deal with nothing changed between the runs, and shows where the calls and probabilities landed on one scale. Nothing is stored, nothing is scored, and the record does not move, so it works on deals whose outcome is already settled, which are the only ones with an answer to compare against. Use it when you want to know whether a judgement is steady before you lean on it.

The lens was being graded twice, once inside the analysis and once by the rubric-backed reader on the Decision tab, and the two could disagree by a grade. Because the fit verdict takes the worst grade, that was enough to show "a fit" on one tab and "a developing fit" on another. There is now one grader, and it refreshes whenever the analysis is regenerated. A read you have confirmed is never overwritten.

The amount a founder is raising is free text, and until now nothing recorded which currency it was in, so a reader handed "1.5" had to decide for itself whether that meant dollars or yen, thousands or millions. Intake now asks, and where the unit is genuinely unknown the reader is told so plainly instead of guessing. Judgements that turn on the size of a round are steadier for it.

The Technical Advisor, the Diligence Associate, General Counsel and the Market Analyst can now be switched on for a room. The Technical Advisor commissions an outside read on a deal's claims; the Diligence Associate traces every figure in a write-up back to the material it came from and says plainly when it cannot; Counsel reads the structure; the Market Analyst follows what is moving around a position you already hold.

Independent reads could record invest or pass and nothing between, so a genuine "not yet, and here is what would change it" had to be filed as one of the two. Watch is now a call a read can make, and it appears everywhere the other two do.

Longer findings from the staff arrived as an unbroken wall of text. They now render with their structure intact, and anything long enough to bury its point opens with its first sentences and expands on request.

A firm that had renamed its dimensions was told its fit came from a rubric it never chose, directly above a card grading the one it did. The banner now names your dimensions, in your order, in your words.

Both were built-in lists written for one geography, so every firm inherited it in the two fields a partner types into most. They are firm settings now, one per line, and leaving them empty keeps the built-in list.

It read the deck and stopped there. It now takes in the readable documents in the data room and what the company publishes about itself, so the write-up reflects what you actually hold rather than one file.

Opening a deal put you at the top of a long page and left you to find the part that mattered. It now opens on the briefing: what this deal is, what has been read, and what is waiting on you.