Vinod Khosla

Debt markets had been getting a little shaky, and the lesson on the show was to hit the bid and get a deal done when it gets that far down. A lot still has to happen between now and the Anthropic IPO in the middle of November.

The month is close enough to name and still far enough that the remaining steps are the risk. The point of hitting the bid is that waiting through shaky debt markets can take the outcome away.

Listen Labs was an early contender in market research that executed well and reached real revenue, in a use case near the top of what AI and LLMs can do at the application layer. Salesforce has liked many of its acquisitions. The doubt is whether Listen Labs is large enough to count.

Salesforce is coming up on sixty billion dollars, with a sixty billion target, and is already almost at fifty. Twenty million dollars of next-generation AI survey revenue may not move the needle. A deal has to tuck in or be big. A clever product on someone else's platform can be a feature masking as a company, which is why more application companies may sell rather than stay independent.

Sequoia made hundreds of millions of dollars when Salesforce bought Listen Labs for $2 billion. Offers had reportedly been at $1.5 billion. Brian Schreier at Sequoia saw about $850 million back to investors on 96 invested. Sequoia led the seed and the A and made about a 25x.

The category is market research, a large fragmented spend. Qualtrics, Medallia, and SurveyMonkey were the last generation of companies in it, and they reached high single digit or low double digit outcomes. Sequoia had stayed in the deal from the seed.

Eleven Labs doubled its valuation to twenty-two billion dollars. The case for putting up to ten percent of the fund in now is that time is short on an eighteen-month deployment cycle. The lead keeps widening under intense competition. Voice is one of the ten most important underpinnings of agentic applications, growth is fast, and the margins are impressive.

A round aimed around fifty billion could land as revenue approaches a billion. Eleven Labs has built a strong team and has sub millisecond response rates for voice, which has become a key modality. One panelist called the asset highly differentiated.

Beam may be as strong as advertised, but it was not on Open Router yet, so the claim could not be checked before the end of the show. Evals need asterisks and daggers, and too many of them focus on chess championships, which is why Dev Ittycheria stays skeptical. Nemotron, thinking machines, and Reflection sit in the same open source pile.

Asked what share of open source tokens runs through US models versus Chinese models in 12 months, the reply was that developers switch more easily than they switch a database. Half in 12 months is a number Dev Ittycheria could see, and sovereignty plus a US base matters. Parity with the leading models is still unproven, so the split is a guess rather than

Reflection has been talking about shipping for a long time. An open weight US sourced model near frontier quality, even six or twelve months off, is a strategic position because of how the market is set up.

Enterprises can get a US open source model near frontier intelligence at three to four times lower cost. There is hesitation about Chinese open source models when the work touches sensitive information or a regulated industry. The speaker called the announcement noteworthy and disclosed being a co-investor in Reflection.

Vinod Khosla led Factory's series C and the most recent round, then posted on Twitter. The wording was a struggling second tier competitor, an absolute demolition of Matan. On The Twenty Minute VC, the reaction was shock, and the post handed every other firm a weapon when competing with Coastal Ventures: a rival can ask whether Khosla is the partner you want when things go bad.

The post was called a mistake Khosla will probably wish had not happened. A lot of the argument is legal obligations versus the implied rules of venture. Breaking the implied rules still has consequences even when no lawsuit follows.

Chris Degnan, the longtime Snowflake sales leader, was a board observer at Factory and spoke to Matan every day, according to the founder. While still that close, Degnan was interviewing for a role at Cognition and then took the CRO job there. Matan came out and said the move was wrong. On 20VC, one panelist says he knows Matan personally and is an angel investor, and Sequoia is an investor in Factory as well, which is why the split was discussed before the conversation turned to Vinod Khosla.

An advisor with a founder's confidential plans should be clear about leaving, especially when the next job is a competitor. A CEO would be annoyed to watch someone with the plans join a rival. CRO job

Early in Q3 the OpenAI teams were tweeting. July was a killer month. August was a killer month too. They had internalized that Q2 was the first time Anthropic GAAP revenue was larger, at 11 billion versus 6.8 billion, growing 10x versus growing 2x.

A 70% number changes the narrative back no matter what Anthropic prints next, if the figure is real. GAAP revenue growth of 70% quarter on quarter, or even 60% up from 18% between Q1 and Q2, is a massive reacceleration for OpenAI after the forward purchases of compute, data centers, and land.

OpenAI has shown that a billion people using ChatGPT a week is a great distribution channel. Enterprise buyers still sign contracts, train thousands of employees, and build workflows, so a switch inside a company can be harder than a switch in a consumer habit.

The run rate looks even more impressive if OpenAI is already embedded and already through procurement. Even with that hurdle, the change inside enterprises has been fast.

OpenAI is back at a $70 billion revenue run rate. A lot of B2B work only worked on Anthropic at the end of last year, and coding at the start of the year only worked on Anthropic. Summer was when OpenAI was on the rope, because the consumer path was the wrong one and the team could not get back.

The OpenAI team then got back in the game and made models people are switching to because they are better and cheaper. The pace at that scale is the point, and the 70 billion figure says it all. Counting OpenAI out three or four months earlier did not survive October.

Rory O'Driscoll leans to liquidity at the margin between OpenAI at 1.4 trillion and Anthropic going public at 2. Liquidity is a lot better to have than not have liquidity when the amount of information is as unknown as it is. Rory O'Driscoll would much prefer to be in the liquid stock rather than accept holding for three or four years because that is what the risk of being private is.

Rory O'Driscoll recalls the first above a billion dollar round in the mid-90s. Observers thought the size was crazy. The first venture round above a trillion dollars represents moving right along with three orders of magnitude. Rory O'Driscoll believes OpenAI is big enough to go public and the company should p

Two or three years of YC time as a partner is equivalent to 20 years in venture. The 15 partners are practitioners of product market fit. Tom Blomfield created Monzo, worth billions of dollars and one of the top fintech companies in the world. People on the team have sold companies for hundreds of millions of dollars.

The reps are why the organization learns. A partner reads a thousand to two thousand applications every single year and writes 50 seed checks in a given year, sometimes 75. Pulling out the billion dollar idea for founders is hit or miss and requires inspiration. Paul Graham and Vinod Khosla are the standard, and all of the YC partners are in the top two to five percent in the