1. Podcasts
  2. Tech Brew Ride Home
  3. Startup Bios Sampler
Tech Brew Ride Home

7 Nov 2026 · 27m · 15 chapters

In short

The episode is a bonus “Startup Bios Sampler” featuring two startup spotlights. Topic 1: Hone (seed $60M; valuation $285M).

Guests

none interviewed; founder/CEO Moritz Stephan is discussed, plus investors and team members (Elad Gill, Scott Wu, Peter Fenton, Shardul Shah).

Key claims

“AI that owns outcomes,” where engines pursue goals for weeks/months, learn company context, and are judged by profit/loss.

Notable examples

improving customer retention; sales lead conversion; Apollo 13-style simulation; a “director” replay system that reduces breaking changes from 33% to 5% across 125 scenarios. Topic 2: Ghost.

Guest

CEO Zane Javed (19). Background: former quant interest; inspired by ChatGPT; built Ghost with cofounders (Nicholas Chua, early-20s).

Key claims

Core is a $3,499 “brain in a box” running local AI agents with encrypted data, no subscription, and a credential firewall. Examples: proactive health/flu-season notifications; integrations with email/calendar/finances/files and smart home devices.

Written by AI. May contain mistakes.

Chapters

1:02

Introduction to Startup Bios

Overview of the Startup Bios podcast and its focus.

“It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration block, or finally break down that long article you've had open for weeks.”
1:43

Hone's $60 Million Seed Round

Discussion of Hone's recent funding and its innovative approach.

“Like and subscribe to hear about the newest startups coming out of stealth and raising new rounds almost every day.”
3:43

Understanding Hone's AI Engines

Explanation of how Hone's AI engines operate and their potential impact.

“Today, you can buy software that helps a sales team figure out which incoming leads look promising.”
5:50

Challenges of AI and Business Goals

Exploration of the challenges faced by AI in achieving business objectives.

“And the company's pitch is that all of that accumulated knowledge belongs to you, the customer, as your IP, not to some model everybody else is also using.”
8:00

Founder Backgrounds and Experience

Insights into the backgrounds of Hone's founders and their previous ventures.

“So they built an agent they call the director that plays the role of the outside world during a replay, deciding when a person writes back or when a shipment notice shows up.”
12:00

Customer Insights and Market Potential

Discussion on Hone's current customers and market positioning.

“Peter Fenton brought up security and I think he's right to.”
14:20

Introduction to Startup Bios

Get introduced to the Startup Bios podcast and its focus on emerging startups.

“Compatibility and availability varies 18+.”
14:45

Zane Javed's Journey to Ghost

Hear about Zane Javed's background and his transition into the startup world.

“Like and subscribe to hear about the newest startups coming out of stealth and raising new rounds almost every day.”
15:18

The Founding of Ghost

Discover how Ghost was founded and the young team behind it.

“Ghost came out of stealth on Monday after about a year of building, and along with that, it announced an$11 million seed round led by Andreessen Horowitz.”
15:37

Core: The Personal AI Computer

Learn about Core, Ghost's hardware designed for personal AI agents.

“Javade is the CEO and his co-founder, Nicholas Chua, is also 19, and the two other co-founders are in their early 20s.”
Show all 15 chapters
16:28

Specifications and Usage of Core

Explore the technical specifications and intended usage of the Core device.

“There's no screen on it, and you're not plugging in a keyboard or a mouse either.”
17:47

Software and Updates for Core

Understand the software that powers Core and how updates are managed.

“On the software side, Core ships with open models already installed.”
19:16

Privacy and Ownership with Core

Examine Core's approach to privacy and data ownership for users.

“the agents learn about you lives on the device.”
21:06

Competition and Market Analysis

Assess the competitive landscape for Ghost and similar products.

“to outside services through a gateway that Ghost runs.”
23:28

Challenges and Future of Ghost

Discuss the potential challenges Ghost may face as it scales.

“That promise about the box working even if Ghost shuts down is, I'd argue, the most underrated thing in this announcement.”
A man walking through an airport, asking his phone a question

Ask about this episode

Get the gist, explore an idea or ask about something said.

Claude

On Tech Brew Ride Home, what are the key points of this episode?

Searching your listening

Transcript

Automatic transcript. May contain errors.

0:28This episode is brought to you by Google Chrome. Hello from Dubai, everybody. Home of the world's craziest traffic. I will have a better setup on Monday to record the regular episode. But real quickly, I wanted to get out of this bonus episode, give you two samples of my new podcast, Startup Bios. Remember, if you want to hear more of this, search your podcast up now.

0:59This episode is brought to you by Google Chrome. You think you know a browser, but Gemini and Chrome, that's new. It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration block, or finally break down that long article you've had open for weeks. Gemini and Chrome is here for it. Ready to make anything online make sense? There's no place like Chrome. Check responses set up required. Compatibility and availability varies 18+.

1:25Hone, a five-month-old startup building AI engines that take ownership of business goals for weeks or months on end, has raised a$60 million seed led by Benchmark and Index.

1:42Welcome to Startup Bios. I'm Brian McCullough. Like and subscribe to hear about the newest startups coming out of stealth and raising new rounds almost every day. $60 million is a really big number for a seed round and hone is raising it at five months old benchmark and index ventures led the round and the company says it's now valued at 285 million dollars according to bloomberg that valuation figure comes from hone itself by the way and nobody said whether that's pre or post money so i'd hold it a little loosely elad gill and sv angels are in as well and so is Scott Wu, the CEO of Cognition, who invested personally.

2:21Peter Fenton from Benchmark and Shardul Shah from Index are both joining the board. So what is Hone actually building? The CEO Moritz Stephan put it this way to Bloomberg, quote, what we are doing is building AI that owns outcomes, end quote. That's kind of the entire pitch. We've been talking a lot lately about AI agents and most of what's out there is agents that do tasks. You give it a job, it does the job, and then it waits on you for the next one. Hone wants to go a level above that. You don't hand their system a task, you hand it a goal, and the goal might take weeks or even months to hit.

3:01Hone calls these things engines. The example they use on their launch blog is something like improve customer retention. You assign an engine to that goal, and then the way they describe it, the engine onboards itself. It learns the company's systems, the people, the history, and the constraints, so it can figure out what good actually looks like there. Then it decides on its own what needs doing to hit that goal and who it needs to pull in to help inside whatever guardrails you've set. And because it owns a goal instead of a task, Hohn says an engine can come back and propose new things it could take on or new ways of going about the work.

3:42Bloomberg's example was sales. Today, you can buy software that helps a sales team figure out which incoming leads look promising. Hohn's idea is that instead of a tool that scores the leads, you'd have an engine whose job is getting more of those leads to actually convert, and it just keeps working on that proactively over time. Stefan told Bloomberg the technology is meant to, quote, look at what a team does its goals and responsibilities and deploy engines to own these, end quote. And it checks in with a human when it needs to. Hone laid out the bigger philosophy in a kind of manifesto on its website.

4:21AI models keep getting smarter, but the world doesn't really look that different than it did five years ago. And Hone thinks the reason is that today's agents are still basically tools inside workflows that humans run. A person has to break the goal down into tasks, feed the agent contacts and catch it when it screws up. So no matter how good the model gets, you're limited by how much human attention there is to go around. Their analogy for this is electricity. Hohn points out that something like three decades went by between Edison's Dynamo and Ford's assembly line, and the big productivity gains didn't come from making more power, they came from reorganizing the factory around it.

5:02So Hohn's line is, quote, quote, the dynamo is here. At home, we're building the factory, end quote. Instead of telling the AI what to do step by step, you tell it the end state you want and what not to break. Their example is something like, cut drop-off in your onboarding funnel to 8 % and keep the other metrics steady while you're at it. They actually describe it as, quote, less chatbot and more Kubernetes cluster. If you're not a developer, Kubernetes is software that keeps servers in the state you asked for and fixes things on its own when they drift. Hone also wants its engines to stick around.

5:39Most AI right now, the session ends and whatever it learned kind of evaporates. Hone says engines are full-time and permanent and they're supposed to get better the longer they're in a role. And the company's pitch is that all of that accumulated knowledge belongs to you, the customer, as your IP, not to some model everybody else is also using. And how you grade? one of these things? Hohn says the real test is the profit and loss. It's not about how slick a demo is or how many tasks it knocked out. The way they put it in the blog, you judge an engine the same way you judge any serious employee by asking, quote, what did you get done this week?

6:20Now, the obvious worry is that you're giving an AI access to sensitive company systems and letting it run for months with a lot of leeway. Hohn's answer is guardrails. Before it takes on any responsibility and engines' decisions get simulated first. Anything sensitive needs a human to approve it, and everything gets logged so somebody can go back and look. Hone's second blog post goes deep on that simulation piece. It opens with Apollo 13, where engineers on the ground tested their fix in a simulator before they sent the instructions up to the astronauts. Hone's point is that today's AI agents mostly don't do that.

6:57They make a change and find out in production whether it broke something. The thing is, Hone's engines change themselves. They rewrite their own tools, their code, their memories, all as they learn. Any one of those changes can break something that's supposed to work. So before an engine locks a change, it replays it against past situations to see what else it might mess up. Their example goes like this. Imagine a manager named George tells the engine, check with me before you order from a new supplier so the engine writes itself a rule always ask george first but when they replay that rule against past situations it turns out it breaks something there was a case where a production line was down and somebody said order the bearings now and with the new rule the engine would have sat there waiting on george instead so the rule gets fixed too always ask george unless the line is down and it only goes live once it passes the old cases too The hard part, Hone says, is that you can't rewind a company.

7:59You can't roll your Salesforce back two weeks. So they built an agent they call the director that plays the role of the outside world during a replay, deciding when a person writes back or when a shipment notice shows up. Other agents stand in for each software tool, and Hone says those fake tools got the real facts right about 90 % of the time in tests. They also fast-forward the clock so weeks of work can replay in minutes. Hohn claims that across 125 test scenarios, replaying a change first cut the share of changes that broke something from 33 % down to 5%. That is their own internal test, of course, and nobody outside has checked that yet.

8:40The founder Stefan is Austrian, and he's been starting things since high school. Back in Vienna, he and a partner launched a startup called Lobu that delivered books from a small local bookshop by bike as an alternative to Amazon, and they even pitched it on Austrian TV. As a teenager, he also co-founded Tumbleweed, a project to build a wind-powered rover for Mars that came out of a European Space Agency incubator program. It grew to more than 50 people across four countries, and prototypes got tested in desert simulations of Mars missions in Oman and Israel. Then he went to Stanford for computer science and did AI research in Chelsea Finn's lab, and that work got accepted at NER IPS in 2022.

9:23While he was there, he also started an anonymous campus app called Fizz with Carlo Coby, and Fizz went through Y Combinator. After that, Stefan joined Cognition, the company behind AI coding agent Devin, as one of its first employees. He worked on how Devon plans and carries out its work and eventually became chief of staff to Scott Wu. Kobe is one of Hone's co-founders, too. He's a Thiel fellow who left Harvard to start Fizz and ran it for several years. The third co-founder, Oliver Brady, was one of the early engineers at the legal AI startup Harvey and then helped build the enterprise business at Mercur.

9:59As for customers, right now we know of two, Cognition and the AI inference company Modal. with cognition hone is building agents to help with the go-to-market side as cognition grows and shardou shah from index basically made the investment case in one sentence to bloomberg quote what cognition is to software development hone can be to every other function in an organization end quote now the first thing i'd point out is the cognition connection cognition is an early customer cognition's ceo is a personal investor and hone's ceo was his chief of staff That's how a lot of startups land their first customer, so it's not a knock.

10:36But it does mean the one customer we know the most about is the one that knows the founder best. We haven't seen revenue, we haven't seen results from a customer, and the valuation number is the company's own. So for now, this is a bet on a team and an idea. And the idea is a really hard one. A coding task is pretty bounded. The tests either pass or they don't. a business goal like improve retention stays open for months while the market and everything else moves around there was a research paper this past june from ceo bench that had ai models run a fake startup for 500 simulated days doing things like pricing and marketing and budgeting and every model they tested did worse than a simple rule-based baseline that's a simulation of course and it says nothing about home directly but it gives you a sense of how hard long horizon business decisions still are for these models right now.

11:27The other problem with owning an outcome is figuring out who gets the credit when a number moves, the engine or just a good quarter. Hone even lists grading and credit assignment over month-long stretches as one of the problems it's working on, so they clearly know that one's coming. The competition is pretty much everyone. Decagon and Sierra are well-funded and already selling AI agents to businesses. Salesforce has Agent Force and OpenAI keeps pushing deeper into agents for office work, Hone's argument would be that those are mostly task tools and it's going after the layer above them. Peter Fenton brought up security and I think he's right to.

12:07He told Bloomberg that Hone is working on guardrails to protect customers if one of its agents causes a security incident, which is a fair worry after a run of cases where AI tools went rogue. And he said new technology needs quote appropriate immune systems an agent that runs for months with access to your systems is much bigger as a target than a chatbot the replay testing idea looks like a real engineering answer to part of that problem at least on paper though i'd admit so yeah 60 million is a lot of rope for a five-month-old company if hone can get from ai native early adopters like cognition and modal to boring old school big companies and actually show them a number moving on the p &l line this round will probably look cheap if it can't well that's a really expensive way to learn that companies still want a person owning the outcome at least for now that's it for this episode like and subscribe if you'd like to get your founding team interviewed on this podcast for 45 minutes or so get in touch at hi at startup bios dot tech that's hi at startup bios dot tech and we'll schedule something.

13:15And if you'd like to get this in your ears every morning, check out the podcast Startup Bios wherever you get your podcasts.

13:33This episode is brought to you by Google Chrome. You think you know a browser, but Gemini and Chrome, that's new. It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration block, or finally break down that long article you've had open for weeks. Gemini and Chrome is here for it. Ready to make anything online make sense? There's no place like Chrome. Check responses set up required, compatibility and availability varies 18+. This episode is brought to you by Google Chrome. You think you know a browser, but Gemini and Chrome, that's new.

14:05It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration block, or finally break down that long article you've had open for weeks. Gemini and Chrome is here for it. Ready to make anything online make sense? There's no place like Chrome. Check responses set up required. Compatibility and availability varies 18+. Ghost, a startup run by a 19-year-old CEO just raised an$11 million seed led by Andrews and Horwitz to sell Core, a$3 ,500 computer built to run your personal AI agents.

14:44Welcome to Startup Bios. I'm Brian McCullough. Like and subscribe to hear about the newest startups coming out of stealth and raising new rounds almost every day. Zane Javed is 19 years old and he's kind of already had one career. He told TechCrunch that for most of his life, he wanted to be a quant, one of those people who use math and data to make investing decisions. And for a while, he actually was one, but he got bored with it. Then ChatGPT came out in 2022, and Javade told TechCrunch, quote, I felt fundamentally excited in a way that I never had been before, end quote. That got him interested in personal AI, and that interest is how we got a company called Ghost.

15:23Ghost came out of stealth on Monday after about a year of building, and along with that, it announced an$11 million seed round led by Andreessen Horowitz. Abstract, Audacious Ventures, SV Angels, and Nova are in the round as well. Javade is the CEO and his co-founder, Nicholas Chua, is also 19, and the two other co-founders are in their early 20s. So this is a really young team, and what's a little unusual is that they're building actual hardware, which you don't see a lot from founders this age. The product is called Core, and the way Ghost describes it, it's a personal computer designed specifically for AI agents, the kind of agents that go off and take actions on your behalf.

16:04The idea is that you pull all of your stuff into one box, like what's on your desktop, your apps, your smart home devices, and that box runs agents around the clock that act as your assistant, do tasks for you, and they claim actually come to understand your life. Javade told TechCrunch, quote, you can run models on existing hardware, but it's a terrible experience on many levels, end quote. He calls Core a brain in a box. There's no screen on it, and you're not plugging in a keyboard or a mouse either. You talk to it through a phone app or a web app by text or by voice, although the voice part also runs through the app, so you're not actually talking to the box itself.

16:40It's meant to just sit on a shelf somewhere in your house and keep working while you go about your day. and it's a pretty serious machine ghost sightless and nvidia rtx pro 4000 blackwell gpu the small form factor version with 24 gigabytes of memory plus an amd ryzen 5 processor with six cores 64 gigs of ram and a one terabyte ssd all in a stainless steel case javade told tech the gpu is bundled into the price so you're not out buying a graphics card on top of it and that price is$3 ,499. There's no subscription after that, according to Ghost. Pre-orders for the first batch opened Monday, and shipping is supposed to start at the very end of October, with Ghost's site saying October 31st specifically.

17:26You get a 30-day return window and a one-year warranty as well. Silicon Angle reports that first batch has already sold out, apparently in a matter of hours, and Ghost is now taking sign-ups for a second batch. We don't know how many units were in that first batch, though, so I wouldn't read too much into have sold out just yet. The team is building these things themselves, by the way, at their own facility in San Francisco. On the software side, Core ships with open models already installed. TechCrunch lists three, two from Alibaba's Quen 3.8 family plus Google's Gemma 431B, and Ghost's own site lists a fourth one called Muse Glimmer 30B.

18:01You can also pull down other models from Hugging Face or load models you trained yourself. Ghost built its own software layer on top with its own browser and filing system, and the pitch is that it keeps working on tasks without you constantly prompting it. Updates work kind of like your car if your car is a Tesla. Javade told TechCrunch, quote, as new models or improvements to existing models come out, we can push over-the-air updates similar to how Tesla works. If a whole new generation of a model comes out, Ghost says it'll ask you before switching you over, or you can just turn on auto updates.

18:34A lot of people in the AI tinkerer crowd have been buying Mac minis lately and setting them up to run AI tools at home, as I'm pretty sure you know. TechCrunch asked Javade about exactly that. His answer was basically that a Mac mini is a general-purpose computer you have to configure yourself, while Core's hardware and software were built for agents from day one, with the models already on there when you open the box. But the real pitch here is ownership. Ghost's first ever post on X said that an AI agent with an intimate understanding of your life should, quote, not live on infrastructure someone else owns.

19:08So instead of renting an assistant from OpenAI or Google every month, you buy the box once. The models run locally. The memory of everything the agents learn about you lives on the device. And Javade says your personal information is encrypted so that no cloud provider or third party, and that includes Ghost, can meaningfully get at it quote the user owns all of it he told tech crunch once you buy the hardware everything is stored run and processed on the device end quote they're also promising the thing keeps working even if ghost itself goes away javade said quote all of the source code logic and model weights live on the device not the services we build this device to not be dependent on having an outside entity maintaining the service end quote you get to decide how much autonomy your agents have, Javade said.

19:55And because an agent like this is going to be holding your logins and passwords, the team is building a firewall to limit what it can do with those credentials. The way he describes it, the firewall watches every request going out of the computer, and if the agent tries something it shouldn't or tries to sneak data out, the request gets blocked and flagged for you to review. What does it actually hook into? Ghost site lists your email, your calendar, your finances, your files, your browser history, your health records and on and on plus smart home stuff like lights and speakers and cameras even fridges and wearables like whoop aura and eight sleep javade gave tech crunch an example of how it's all supposed to come together quote core proactivity sends you a notification that you might be getting sick because it sees your resting heart rate is elevated sees from computer history that you were up late working on a project and has seasonal context that it is flu season in san francisco the examples on the website are smaller than that like telling you you're low on groceries or to leave early for your commute because of the weather the privacy promise does have some fine print though ghost's home page says quote no data ever leaves your home but its own support page says web search email and those over-the-air updates need an internet connection and its privacy policy says those online features can send task information out to outside services through a gateway that Ghost runs.

21:15Ghost says that gateway doesn't store anything, but the outside provider on the other end plays by its own rules. The policy also says, does not use a remote AI model as a fallback for its local conversational AI or memory processing. So the way I read it, the thinking and the memory stays in the box. But when the agent has to go do something out on the internet, some of that task information is going to leave your house, which I I mean, it kind of has to. Javade says he's not too worried about the competition, including the giant tech companies. He told TechCrunch, quote, People sometimes give incumbents credit for having already built the product because they have the ingredients.

21:52He gave Apple credit for great hardware. But he thinks most of the hard part here is software. Quote, The model reasons over your life, how it remembers, how it takes action, how it knows when to reach out. is$3 ,499 a lot compared to a$20 a month chatbot plan. You'd be paying that subscription for well over a decade before you broke even. But I don't think that's really who this is for, at least not yet. The first buyers are going to be the same people who've been buying Mac minis to run agents at home, people who would otherwise be specking out a GPU rig themselves. For that crowd, a pre-configured box with a decent NVIDIA card in it and the software already set up is a pretty reasonable deal.

22:35The thing I'd keep in mind is that a model of around 30 billion parameters running at home is good, but it's still going to trail the biggest cloud models on the hard stuff like complicated reasoning or coding right now. Anyway, Ghost's whole bet is that for a personal assistant, knowing your life matters more than raw smarts. That might be true, but the cloud models keep getting better too, and so do their memory features. The competition is basically everybody. There's Apple, which already sells the Mac Mini that Ghost keeps getting compared to, and which has been leaning hard on its own privacy story with AI.

23:09There's NVIDIA, which sells its own little desktop AI boxes to developers. And there's every big lab pushing cloud agents that are free or close to it. Javade's point that having the ingredients isn't the same as having the product is fair, I think, but Apple could in theory ship a version of this, and the day they do, Ghost's job gets a lot harder. That promise about the box working even if Ghost shuts down is, I'd argue, the most underrated thing in this announcement. Remember Humane? When HP bought Humane's assets last year, the AI pins basically stopped working and the people who paid hundreds of dollars for them ended up with a paperweight.

23:44So putting the weights and the logic on the device and saying so up front is a smart way to take that fear off the table. There are real risks here, though. Building units in-house in San Francisco works for a first batch, but it's a whole different business at scale, especially with GPU prices being what they are. Then there's security. Put someone's bank data, health records, and home cameras into one always-on box, and that box becomes a pretty attractive target. The firewall sounds good, but as far as I know, no outside researcher has actually tested it yet. And the no data ever leaves your home line on the homepage versus what the privacy policy says, that's exactly the kind of gap that can turn into a bad news cycle for a privacy-first company if they're not careful.

24:24Still, Andreessen leading an$11 million seed for a 19-year-old's hardware company tells you they buy the idea that personal AI eventually moves into your house the way the Wi-Fi router did. The real test is the end of October when people actually plug these in and find out if it feels like having an assistant that knows them or just an expensive box on the shelf. That's it for this episode. Like and subscribe. If you'd like to get your startup team interviewed on this podcast for about a 45-minute episode. Get in touch at hihi at startupbios.tech. And if you'd like to get this in your ears every morning, check out the podcast, Startup Bios, wherever you get your podcasts.

25:14All new holiday drinks are now at McDonald's. Like the Frosted Cranberry Splash with Sprite, the Red Bull Frosted Cranberry Energizer, and cookie butter coffees. Tis the season for seasonal drinks at McDonald's.

From the publisher

Search: Startup Bios in your podcast app and subscribe!
Learn more about your ad choices. Visit megaphone.fm/adchoices

More from Tech Brew Ride Home

All episodes