
Cofounder/CEO @HiHello 👋 Cofounder @CartaInc 🥧 Chief Firestarter @K9Ventures 🦮 OG Pre-Seed Investor @lyft @twilio @auth0 @lucidchart @everlaw @forethought_ai
Palo Alto, CA
Joined April 2008
- Tweets30.9K
- Following2.9K
- Followers33K
- Likes43.8K
Dr. Manu Kumar 👋🏽 retweeted
People seem to forget how unprecedented this is
Anthropic has filed an S-1 registration statement with the SEC related to its proposed IPO.
Read more: anthropic.com/news/s1-sec
Legal disclaimer: A registration statement relating to these securities has been filed with the SEC but has not yet become effective. These securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post does not constitute an offer to sell or the solicitation of an offer to buy these securities, nor shall there be any sale of these securities in any state or jurisdiction in which such offer, solicitation or sale would be unlawful prior to registration or qualification under the securities laws of any such state or jurisdiction.
Dr. Manu Kumar 👋🏽 retweeted
Amazon Zoox robotaxi, a Google Waymo car, and a Tesla Cybercab became trapped in a viral face-to-face standoff on a narrow city street in 🇺🇸Austin, Texas, showcasing a classic real-world artificial intelligence deadlock.
The vehicles met head-on on a narrow side street constricted by cars parked on both sides. Because each company programs its vehicles with proprietary decision-making algorithms, the vehicles struggled to establish right-of-way protocol without a human driver to initiate a courtesy gesture or negotiate positioning.
Eyewitnesses noted that the Waymo continued to creep aggressively forward instead of yielding. Meanwhile, the Tesla Cybercab - which runs entirely unsupervised on vision-based software without a steering wheel or pedals - was seen reversing to clear space for the others.
Observers also pointed out that Zoox’s unique bidirectional carriage design makes native reversing straightforward, though both it and the Waymo remained locked in position for several minutes before traffic could clear.
This event highlights a prominent operational barrier for autonomous fleets: multi-platform coordination. While current robotaxis excel at perception and safety when interacting with fixed obstacles or rule-following humans, they struggle when they encounter another AI entity governed by a completely different set of non-standardized routing protocols. Similar standoffs between Zoox and Waymo vehicles have previously been documented in San Francisco, indicating that “algorithm clashes” are becoming a frequent consequence of multi-fleet urban deployments.
Can the robotaxi companies program “courtesy gestures” as part of their algorithms?
Dr. Manu Kumar 👋🏽 retweeted
Anthropic files for IPO. But who is poised to get the 💰? My usual breakdown below 👇
Investors will be sharing nearly $1 TRILLION of proceeds on ~$142b invested. A ~6.5x blended multiple on a huge amount of capital, in just 5 years. 🤯
Note: all those numbers assume a $1.5 trillion IPO valuation.
So let’s dive in 🔎
1) The craziest fact: no VC in the Series A or Series B 🤯
The $124m Series A in May-21 was led by Jaan Tallinn (Skype’s co-founder), with Dustin Moskovitz and Eric Schmidt. It is likely to return ~345x the original capital, or ~$43b to those early angels. Mind-boggling.
The $982m Series B in 2022 was led by... SBF and FTX: it will return ~79x the capital invested. FTX’s $500m would be worth ~$39b today.
Plot twist: the estate sold it for ~$1.3b in 2024, in one of the worst trades ever.
2) The first VCs in: Spark and Menlo are killing it 🏆
Spark will make over $10b from its Anthropic investments.
They first came into the 2023 Series C and are making ~79x their investment in that round. They followed in every single round since, investing $777m in total, now worth ~$11b.
Menlo led the Jan-24 Series D and will make ~30x its money on that round. And finally, Lightspeed’s Series E lead will make ~16x in 18 months.
3) The real winners in $ terms: the strategics 💰
Amazon put a grand total of ~$18b in, mostly via convertible notes converting at ~$2 a share, against $58.90 for the Series H investors.
That gives Amazon ~21% of the company, worth ~$310b. That’s ~17x its investment. 💵
Google is not that far behind: its notes and the $10b it put in this April alone will be worth ~$83b.
And that’s before the compute: Anthropic has committed $110b+ to AWS and $111b to Google Cloud. They are both investors and suppliers, so they win twice.
4) Founders and team: incredible outcome, and giving most of it away 🙏
The seven co-founders hold ~240m shares each. That’s ~$21b per person, or ~$150b combined. 👏
And each of them has pledged 80% of it to philanthropy. So ~$120b will be going to good causes. On top of that, Anthropic put 5% of the company (~$75b) into its own Public Benefit Fund.
And the IPO will mint a huge number of UHNW in the team: they will share ~$355b worth of shares, options and RSUs. Perhaps the biggest and fastest employee wealth creation event in history.
Congrats to Dario, Daniela and everyone involved. Legendary. 👏
Dr. Manu Kumar 👋🏽 retweeted
Tesla vs Waymo driverless miles in Austin, TX
Via @RtaxiTracker
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.
Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.
Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.
But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?
We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.
Read the full report: cohere.com/blog/concerning-b…
Dr. Manu Kumar 👋🏽 retweeted
Right now in the US, part of the Democratic Party has decided to play the "socialist" card. Days before the midterms, DSA candidates are winning primaries and the word isn't taboo anymore.
So today I'm going to show you why socialism never works. Not with ideology. With mechanics.
Let's start with a dumb question. How much should a loaf of bread cost?
You don't know. Neither do I. Nobody does. And that's exactly the point.
The price of a loaf of bread is the instant synthesis of millions of pieces of information that no single person holds: the wheat harvest in Kansas, the price of gas for the oven, the baker's wages, the shop's rent, a drought in Argentina, a war on the Black Sea, whether people feel like bread this morning or a bagel.
In 1920, Ludwig von Mises published a paper that should have ended the debate. His argument: without private ownership of the means of production, those means are never exchanged. No exchange, no prices. No prices, no economic calculation. The planner can't know whether a steel mill is worth more than a dam or a hospital. He's flying blind. He can't even tell that he's wrong.
In 1945, Friedrich Hayek drove the point home in "The Use of Knowledge in Society." Useful knowledge is scattered across millions of minds. The baker knows things about his neighborhood that no ministry ever will. Prices are a telecommunications system: when tin becomes scarce somewhere in the world, its price rises, and millions of users economize on it without even knowing why. No committee, no ministry, no plan can replicate that. The market isn't an ideology. It's a distributed information processor.
Then Milton Friedman picked up a simple pencil. The wood comes from Washington State, the graphite from South America, the rubber from Malaysia. Thousands of people who don't know each other, don't speak the same language, maybe even hate each other, cooperate to make something none of them could make alone. Nobody gave them orders. Prices were enough.
Now run the experiment in reverse. Set the price yourself.
Cap rents below market: landlords stop maintaining, stop building, housing disappears. Swedish economist Assar Lindbeck, a man of the left, said rent control is the most efficient technique known for destroying a city, short of bombing it.
Cap the price of flour: shelves empty, lines form, the black market takes over. Venezuela, 2016. The country with the largest oil reserves on earth, and people standing in line for hours to eat. Nearly 8 million Venezuelans eventually fled.
Socialism smashes the thermometer and then acts surprised that the fever keeps rising.
And this is where it gets serious.
Because when the plan fails, and it always fails, the planner has two options. Admit the plan is wrong. Or blame the people. Saboteurs, speculators, kulaks, enemies of the people.
He always picks the second one.
Hayek saw it in 1944 in The Road to Serfdom, in a chapter with a chilling title: Why the Worst Get on Top. A central plan requires central power. Central power attracts those who crave power. And when reality resists the plan, reality has to be broken. Which means people have to be broken.
Stalin. He collectivizes the land. Output collapses. He blames the peasants and seizes the harvests. The Holodomor, 1932-1933: millions starved to death in Ukraine, the breadbasket of Europe. Add the Gulag, the purges, the deportations.
Mao. The Great Leap Forward, 1958-1962. Backyard furnaces producing useless steel, insane grain quotas, local officials inflating numbers to avoid punishment, and the state requisitioning food based on those fake numbers. The result: the worst famine in human history. Estimates range from 30 to 45 million dead.
Pol Pot. The Khmer Rouge abolish money, markets, property, and empty the cities. The dream of total planning, carried all the way. In four years, roughly one in four Cambodians dies.
Hitler. He sent communists to the camps, but his party called itself National Socialist and the economy answered to the plan and the party. Companies stayed private on paper, but the state decided what to produce, at what price, and for whom. Hayek dedicated The Road to Serfdom "to the socialists of all parties" precisely because he saw the same root: the idea that the state must steer society toward a collective goal, and the individual must step aside.
Different eras. Different continents. Different cultures. Same mechanism. Same ending.
This isn't a bug. It's a feature of the system. Remove prices and you remove information. Remove information and you remove the ability to correct. And when you can no longer correct through prices, you correct through force.
So let me say it plainly. Socialism always starts nice. A rent freeze. Public grocery stores. Make "the rich" pay. Nobody promises the Gulag on day one.
But if America truly goes down this road, the rest of the story is already written. The plan will fail. There will need to be culprits. There will need to be power to punish them. And a socialist government holding the American state apparatus, its surveillance capabilities, its technology, its military, wouldn't be a regional tragedy. It could be the greatest tragedy in history.
The 20th century paid for this lesson with tens of millions of lives. We don't get to forget it in one generation.
Prices aren't the enemy of the people. Prices are the voice of the people.
Dr. Manu Kumar 👋🏽 retweeted
There’s a world of difference between user-authorized agents (Muse, Instinct, Grokbot, etc.) and true unauthorized/rogue agents that could constitute cyberattacks.
Most of the activity described in the Axios story appears to be the former: customers sharing credentials with agents so they can take actions on their behalf. This may violate bank terms requiring humans to log in themselves, but these are in no way cyberattacks.
The direction banks should go is to support user-directed agents. For example, Amex is an authorized connection in Muse, making it easy for users to grant, limit, and revoke access.
Read-only permission levels are a sensible interim step. Several banks already support or are moving toward them. Full write access (with user credentials) is a different risk category and should be approached carefully.
Frightening the public by blurring the line between user-authorized agentic behavior and cyberattacks is irresponsible.
SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios.
The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios.
Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said.
Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior.
Read my latest for Axios here: axios.com/2026/10/22/ai-agen…
Dr. Manu Kumar 👋🏽 retweeted
Disaster for pre-IPO founders in California.
Replying to @RoKhanna
I absolutely cannot afford to stay, given that you and your corrupt union overlords are planning another "one-time" tax in two years. This is not a 5% tax, it's a full confiscation of all my liquid assets. It's going to bankrupt founders without full liquidity.
Dr. Manu Kumar 👋🏽 retweeted
Replying to @RoKhanna
I absolutely cannot afford to stay, given that you and your corrupt union overlords are planning another "one-time" tax in two years. This is not a 5% tax, it's a full confiscation of all my liquid assets. It's going to bankrupt founders without full liquidity.
Effective Immediately, all organizations using PHP for their systems are advised to be prepared for an update addressing two critical vulnerabilities. PHP is widely used for web applications. These vulnerabilities often provide a direct path to the server's file system or database. If an immediate update is not feasible, it is recommend that organizations take the system offline until it can be fully mitigated. NSA will follow up shortly with more specific guidance.
Dr. Manu Kumar 👋🏽 retweeted
Lots of VCs kindof sound like this: "Your basketball team doesn't have a moat, any healthy human can throw the ball into the basket thingy. "
People who have not run companies before don't understand how much the team matters more than some imagined structural moat.
Dr. Manu Kumar 👋🏽 retweeted
About pre-seed/seed
Real Madrid can buy almost any established star in the world. That doesn’t mean they are automatically the best academy for discovering a 15-year-old before anyone knows his name. Those are different muscles.
Pre-seed and seed aren’t Series A with smaller checks.
But calling ourselves specialists isn’t enough. Only the best will survive!
Categorically vote No on every new city and state tax on the ballot in California.
There is no funding shortage. It’s just fraud.
Dr. Manu Kumar 👋🏽 retweeted
Saved you some time!
Here's how I filled out my 2026 California General Election ballot:
Governor: Steve Hilton
Lieutenant Governor: Gloria Romero
Secretary of State: Don Wagner
Controller: Herb Morgan
Treasurer: Jennifer Hawks
Attorney General: Michael Gates
Insurance Commissioner: Ben Allen
Superintendent of Public Instruction: Sonja Shaw
Prop 1 ❌
Prop 2 ❌
Prop 3 ❌
Prop 4 ❌
Prop 5 ❌
Prop 37 ❌
Prop 38 ✅
Prop 39 ✅
Prop 40 ❌
Prop 41 ✅
Prop 42 ✅
Prop 43 ✅
Prop 44 ❌
Don’t wait for Election Day! Drop off your ballot NOW and spend the next two weeks getting everyone you know to do the same.
Dr. Manu Kumar 👋🏽 retweeted
I was born in Paris. And it pains me to see how decades of delusional policies and tragically upside-down incentive systems rewarding losers instead of good people are now resulting in what can only be seen as the fall of the Republic.
Dr. Manu Kumar 👋🏽 retweeted
Dior had no faith in the French police to protect their store so they brought in their own security team to protect the flagship store in Paris tonight
Many of the super high valuation early stage rounds that are announced have 2-tiered structures. Large firms have adopted the strategy to artificially inflate valuations and the perception of success and traction (king-making) and at the moment it is working for founders. For example, we will invest $10M at $100M and $50M at $1B. They are buying 15% ownership with $6OM, so the effective valuation is $400M, but the announced round is at $1B. That is all fine, except the lead is investing at $400M while the existing investors are asked to do their pro rata and invest at $1B. It will be interesting to see how this plays out over time.
American capitalism’s quality of life is so high that a sitting Senator considers a hot sauce merger to be a matter of national importance.
Grok Voice 3.0 is now the world’s leading conversational voice model
SpaceXAI’s Grok Voice 3.0 is the new leading Speech to Speech model, debuting at #1 on the Artificial Analysis Speech to Speech Index at 86.9, while also being competitive on latency
Grok Voice 3.0 is @SpaceXSI 's successor to Grok Voice 2.0, a Speech to Speech model available through the SpaceXAI Realtime API at $0.08 per minute. We evaluated Grok Voice 3.0 with the Rex voice through the SpaceXAI Realtime API.
Key takeaways:
➤ Speech to Speech Index: Grok Voice 3.0 debuts at #1 at 86.9, ahead of GPT-Live-1 (Astra, medium) at 83.2, Gemini 3.8 Live Extended Thinking (High) at 82.6 and GPT-Live-1 (Sol, low) at 80.1. It is up on Grok Voice 2.0 at 81.3 (+5.6 points) and Grok Voice 1.0 at 73.7 (+13.2 points), released just under six months ago. The Index averages Speech Reasoning (Big Bench Audio), Agentic Performance (Tau Voice), Arena Preference, and Task Success Rate
➤ Speech Agent Arena: Grok Voice 3.0 ranks #3 in preference at Elo 1070, behind Gemini 3.1 Flash Live Minimal (1100) and Gemini 3.8 Live (1088) and ahead of GPT-Live-1 (Sol, low) at 1057, and #1 on Task Success Rate at 94.8%, highest on the board, just ahead of Grok Voice 2.0 at 94.6%. Grok Voice 2.0 trails at Elo 1011
➤ Tau Voice: Grok Voice 3.0 takes the #2 spot on our Tau Voice benchmark implementation at 71.6%, behind GPT-Live-1 (Astra, medium) at 74.5% and ahead of Gemini 3.8 Live Extended Thinking (High) at 68.6% and GPT-Live-1 (Sol, low) at 59.3% - up from 56.5% for Grok Voice 2.0
➤ Big Bench Audio: Grok Voice 3.0 scores 98.5% on audio reasoning, ahead of Gemini 3.8 Live Extended Thinking (High) at 97.7% and Grok Voice 2.0 at 97.2%
➤ Speed: Average Time to First Audio on Big Bench Audio is 0.81 seconds for Grok Voice 3.0, well ahead of GPT-Live-1 (Sol, low) at 1.24s, GPT-Live-1 (Astra, medium) at 1.34s and Gemini 3.8 Live Extended Thinking (High) at 1.35s, though slower than Grok Voice 2.0 at 0.70s
➤ Cost: Grok Voice 3.0 costs $4.80 per hour of input audio, the same as Grok Voice 2.0. That is above GPT-Live-1 (Sol, low) at $4.47 and Gemini 3.8 Live Extended Thinking (High) at $3.50, cheaper than GPT-Live-1 (Astra, medium) at $5.83, and ~2.2x cheaper than GPT-Realtime-2.1 High at $10.75
Congrats to @elonmusk and the @SpaceXSI team 🚀
See below for more detail ⬇️




























