1:04:40Greg Isenberg · Interview
Stop using Claude. Start using Codex?
A 64-minute masterclass where a Codex true believer converts a Claude Code skeptic, live, on camera.
April 27thThe argument in one line.
Autoresearch shifts the human role from tester to approver: you define what better means, the agent runs hundreds of experiments overnight, and you wake up to the winning configuration.
Read if. Skip if.
READ IF YOU ARE…
- You build or want to build SaaS tools and are looking for a moat that compounds automatically.
- You run an agency and want a pitch that out-tests competitors by an order of magnitude.
- You are curious about Andrej Karpathy's recent open-source work and want a plain-English explainer.
- You want concrete starter ideas for AI-powered businesses that do not require a large team.
- You have heard of Autoresearch on social media and want to understand if and how to get started.
SKIP IF…
- You already run a GPU lab and are deep in ML research — this is a builder-audience primer, not a technical deep-dive.
- You are looking for a tutorial with actual code; this is an ideas and mental-model episode.
The full version, fast.
Autoresearch is Karpathy's open-source AI agent that takes a goal, plans experiments, edits and trains on a GPU, reads the metrics, discards failures, and loops until it finds improvements. The core insight is that the human sets the objective and approves the winner — the agent does all the in-between testing. Built on that loop, ten business categories emerge: niche optimization products, marketing A/B engines, research-as-a-service, SaaS power-tool upsells, high-volume testing agencies, trading backtests, CRM lead scorers, finance ops automation, internal productivity labs, and done-for-you due diligence shops. You need an NVIDIA GPU to run it locally, but Google Colab with a free T4 runtime is a viable entry point.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 00:45
01 · Cold open
Name-drop hook: Karpathy + Tobi Lutke. Promise to explain, give use cases, and share starter ideas.
00:45 – 02:40
02 · How Autoresearch actually works
The core loop: set goal, agent plans experiment, edits Python, runs short GPU training, reads metrics, saves improvements, repeats.
02:40 – 03:40
03 · Visual walkthrough of the loop
Hand-drawn flowchart: set-goal to plan to edit/train to read metrics to is-it-better to save/discard to plan-again.
03:40 – 05:14
04 · Mental model: your research bot
Write a clear task, give the bot code/GPU/internet access, bot runs a plan-act-read-update loop, you come back later and review logs, charts, and a written summary.
05:14 – 06:48
05 · Sponsor / free workshop promo
Greg promotes a free live event on building businesses in the age of AI. QR code shown.
06:48 – 08:28
06 · Idea 1: Niche agent-in-a-box products
Package tiny Autoresearch loops for one painful niche. Monthly subscription. Value prop: runs 24/7, shows you the winner to click accept.
08:28 – 09:25
07 · Idea 2: A/B testing for marketing
Auto-test headline/layout/offer variants on landing pages and ad creatives. Sell as always-on experiment engine retainer.
09:25 – 10:43
08 · Idea 3: Research as a service
Point the loop at market/competitor research, investor/M&A due diligence, compliance tracking. Charge per report or monthly subscription.
10:43 – 11:49
09 · Idea 4: Power tool inside your SaaS
Embed an Autoresearch-style optimize button in an existing product. Tune prompts, pick best pricing, rank suppliers. Charge higher tiers.
11:49 – 13:05
10 · Idea 5: Agency that runs 100x more tests
Simple pitch: 100x more testing for same or lower fee. Niches: Shopify conversion lab, B2B SaaS pricing, email optimizer. Revenue share model.
13:05 – 14:24
11 · Idea 6: Auto quant for trading ideas
Run small fast backtests of simple trading rules on one GPU overnight. Keep promising strategies. Trade own account or sell signals.
14:24 – 15:21
12 · Idea 7: Always-on lead qualification
Point agent at CRM and inbound leads. Auto-grades, suggests next actions, drafts follow-ups. Salespeople focus on high-value deals.
15:21 – 16:09
13 · Idea 8: Finance ops autopilot
Loop through invoice matching, expense report generation, exception detection. Sell as software or ops service.
16:09 – 16:53
14 · Idea 9: Internal productivity lab
Treat your own company like Karpathy's GPU lab. Define KPIs, let agents iterate on workflows and templates. Fewer meetings, less grunt work.
16:53 – 17:41
15 · Idea 10: Done-for-you research and DD shop
Use the research loop to chew through docs, filings, product pages, reviews. Living memo for investors/acquirers. Fast structured briefs and monthly update packs.
17:41 – 19:47
16 · Non-business use cases
Clinical trial design is itself a hyperparameter search. Agent swarms could optimize treatment protocols on small proxy experiments before moving to human trials.
19:47 – 21:50
17 · AgentHub announcement
AgentHub: GitHub for agents. Agent-swarm collaboration platform, no main branch/PRs/merges. Greg: watching him speedrun a one-man billion-dollar company.
21:50 – 23:41
18 · How to get started
Requires NVIDIA GPU. No M-series Mac support. Cloud options: Lambda Labs, Vast AI, RunPod, Google Colab T4 runtime. Fastest path: paste GitHub repo into Claude Code, paste commands into Colab.
23:41 – 24:21
19 · Final thoughts
Encourages tinkering early while the tool is still confusing to most people. Plug for ideabrowser.com.
Lines worth screenshotting.
- Autoresearch makes the human the approver, not the tester — the agent runs hundreds of experiments so you only touch the winner.
- You need an NVIDIA GPU to run Autoresearch; MacBook M-series chips are not supported, but Google Colab's free T4 runtime is a practical substitute.
- The fastest way to get started is to paste the Autoresearch GitHub repo link into Claude Code and let it walk you through the installation.
- AgentHub is Karpathy's companion launch: a Git-based collaboration platform for agent swarms with no main branch, no PRs, just a sprawling DAG of commits.
- The agency pitch built on Autoresearch is disarmingly simple: we run 100x more tests than other shops for the same or lower fee.
- Clinical trial design is itself a hyperparameter search — agent swarms could optimize treatment protocols on small proxy experiments and cut trial costs significantly.
- Early movers who tinker with a tool while it has 25,000 GitHub stars and is still confusing to most people build a durable advantage.
- The always-on experiment engine retainer model is structurally more defensible than project-based consulting because the value compounds with each iteration cycle.
- Finance ops automation (invoice matching, exception detection) is one of the highest-value Autoresearch applications because the ROI is immediately measurable in hours saved.
- Treating your own company like a GPU lab — defining KPIs and letting agents iterate on workflows — is the internal version of the same loop every external product uses.
Let the loop run while you decide.
WHAT TO LEARN
Autoresearch does not replace judgment — it eliminates the manual iteration between decisions, so the person who defines goals clearly wins faster than the person who executes carefully.
02How Autoresearch actually works
- The human role in an Autoresearch loop is to define what better means and to accept or reject the winner — everything in between is delegated to the agent.
04Mental model: your research bot
- You do not need to own an NVIDIA GPU to start; cloud rentals like Google Colab's free T4 runtime make the barrier to entry effectively zero.
06Ideas 1 through 10: product and agency plays
- The most defensible business built on Autoresearch is one where more experiments compound into better data, not just better outputs — each loop cycle builds an asset the competitor cannot easily replicate.
- The agency pitch that Autoresearch enables is structurally different from traditional consulting: the value proposition is volume of tests, not hours of expertise, which shifts competition to infrastructure rather than reputation.
11Non-business use cases
- Non-commercial applications — drug trial design, scientific research, financial modeling — may ultimately generate more value than the business ideas, and builders who understand the loop early will have a head start when those use cases formalize.
13How to get started
- Tinkering with a tool while it is still confusing to most people is itself a strategy: confusion is the moat, and the window closes as tutorials accumulate.
- The fastest on-ramp to any new technical tool is to hand its documentation to an AI coding assistant and ask it to walk you through installation step by step.
Terms worth knowing.
- Autoresearch
- An open-source AI agent by Andrej Karpathy that takes a goal, plans experiments, edits and runs short GPU training jobs, reads the results, and loops until it finds improvements — all without human intervention between cycles.
- AgentHub
- A companion open-source project by Karpathy: a stripped-down Git repository and message board designed for swarms of AI agents collaborating on the same codebase, with no main branch, no pull requests, and no merges.
- H100
- NVIDIA's high-end data-center GPU; the hardware Autoresearch was tested on. Consumer and cloud GPUs (T4, A100) also work but may be slower.
- Google Colab
- A free cloud notebook environment by Google that provides access to NVIDIA GPUs (including the T4) via a browser, making it the lowest-barrier entry point for running Autoresearch without owning hardware.
- Hyperparameter search
- The process of systematically trying different configuration values for a machine learning model to find the combination that maximizes a target metric. Autoresearch automates this process.
- ROAS
- Return on Ad Spend — revenue generated per dollar of advertising spend. One of the marketing metrics Autoresearch-powered A/B testing engines are designed to maximize.
- CAC
- Customer Acquisition Cost — the total spend required to acquire one new customer. Lowering CAC is a primary goal of AI-driven ad and landing page optimization loops.
Things they pointed at.
Lines you could clip.
03:15
“Think of Auto Research as a research bot that runs experiments for you while you sleep, tries lots of ideas fast, and keeps the winners.”
11:11
“We do a 100 times more testing than other shops for the same or lower fee.”
18:47
“I'm watching him speedrun a one-man billion-dollar company.”
23:43
“In the fog, people don't really understand where the opportunity is — that's sometimes when there's an opportunity.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
metaphor
00:00Andre Karpathy. I mean, one of the godfathers AI has just launched something called auto research. And auto research is a huge deal, and it's going viral on Twitter.
00:10And I just wanted to do an episode where I can explain to you in the clearest way possible what it is, what are the use cases, how to make money from it, how to be more productive with it, how to create impact with it. And by the end of this episode, I'm gonna give you a bunch of different ideas, use cases for how to use auto research.
00:28I'm gonna explain it to you in the most clear way possible. And at the end, I'm gonna tell you how you can actually get started with it. Um, so let's go right into it.
00:45So what is auto research? Well, it's like having a without you doing the boring stuff. I mean, sounds intriguing.
00:57Right? So how do you actually, you know, program it or get started with it?
01:01Well, first thing is you gotta give it a goal. So you can say something like, make this small AI model smarter. That's the goal.
01:09And then an AI agent will actually plan what to do, like different settings, code changes, edits the Python code for you, runs a short training experiment on a GPU, um, for about five minutes.
01:22It reads the results, it and then it decides what to change next and to repeat the loop. So in some ways, you know, if you've seen my video on the Ralph loop where it basically would do engineering twenty four seven and you'd wake up to new stuff happening, in simplest terms, that's what auto research is helping you you know, it do.
01:43You give it a goal, the AI agent does a thing, you know. You tell the AI what better means, uh, cheaper leads, more clicks, higher sales, better model school, and then the AI keeps changing things, testing them, and it only saves the changes that improve.
01:58So what's really cool about it is you wake up, you grab the best version, and then hopefully you turn it into something you charge for or, you know, you give it away. I saw this tweet by Toby, who's the, uh, CEO and co founder of Shopify.
02:12Auto research works even better for optimizing any piece of software. Make an auto folder, add a program MD, that's just a markdown file, which is really the foundation of what'll you know, how you're gonna be using Auto Research and a bench script, make a branch, and let it rip.
02:28So that's why I started paying attention to Auto Research. Right? When Andre Carpathi, Legend, and Toby's and and and more people, you know, start playing with it, I'm like, okay, I gotta pay attention.
02:39So I created this little visual for for how to think about what auto research is. So you set the goal. Uh, the the AI plant is an experiment.
02:48It edits and trains the code and settings. It runs a short training on a GPU.
02:54By the way, this is an important I should I should mention that you need a a NVIDIA chip to actually run auto research, or you can do it in the cloud.
03:07I'll talk about this at the end of the episode. But you you know, you do need that. You can't just run it on, let's say, have a MacBook m one or something like that.
03:15It reads metrics. It says, is it a better result? If it's if it's not, it's gonna log the attempt and it's gonna discard the config.
03:23If it's yes, it saves it to the config, um, and then just plans a different experiment. And it just, you know, hopefully gets better on your goal, whatever it is.
03:31So let's let's get into we're gonna get into some of the ideas, business ideas around it.
03:40But right before that, I just wanna say, here's a simple mental model for how I'm thinking about auto auto research.
03:47So imagine you have a . Number one, you write, you know, a clear task.
03:54So for code experiments, maybe it's improve this model test score. For business, figure out the top five competitors for product x y z and make a short report.
04:03Step two is you give the, uh, you give the bot, um, you know, access to the code, a GPU for ML experiments. You obviously need to give it access to the Internet and documents if you're doing reading task. The bot then runs a loop.
04:16So it it plans, it acts, meaning it might run code or search, it reads results, it updates the plan. And then you just come back later, you know, uh, it could be twelve hours, twenty hours, six hours, and you see if it's logged everything, charts and metrics, and then it gives you a written sum summary in normal language.
04:35So, you know, think of Auto Research as a research bot that runs experiments for you while you sleep, tries lots of ideas fast, and keeps the winners. Quick break to invite you to something. Now, this isn't an ad.
04:45I just wanna invite you to a free event because I think that you're gonna get a lot out of it. I wanted to take one hour of time where we just talk about building businesses in the age of AI. People say SaaS is dying.
04:57I actually believe the quite opposite. I think that SaaS is just evolving. I think right now is an incredible time to be building software startups that help you craft your dream life.
05:07And for all those reasons, I'm I said, let's just book one hour of time. It's gonna be 11AM, March 12.
05:15That's a Thursday where we can go and lock in and just talk about building businesses in AJAI. I'll include a link in the description in the show notes to join, and I can't wait to see you there.
05:27Okay. How do we use it? Here are some ideas for you.
05:30So the first idea for you I have is a niche agent in a box, you know, products. This can be multiple products. And by the way, I put out these ideas.
05:38I want you to do these ideas. I think that, you know, even if they don't turn into businesses, you will learn about these tools, and that is going to help you outperform 99.9% of people on this planet.
05:50So you package tiny auto research loops tuned for one painful niche.
05:56So the example I think of is an Amazon listing experimenter, an email sequence tuner for real real realtors, a pricing optimizer for SaaS.
06:06Those are, you know, auto research loops and and and and ideally in a niche that you understand well. And then you charge a monthly fee. So the value prop is this thing runs experiments for you twenty four seven and just show shows you the winner to click accept.
06:21How valuable is that? And how many different niches are there that, you know, this plays into? The hard part is figuring out what the what what's the pain points and then and then obviously, you know, you wanna be quick quick to market.
06:33Right? So here's a visual of it. Pick the painful niche.
06:37Design the tiny auto research loop, run experiments automatically, see which setup works best, turn best setup to a simple agent product, and then you charge that monthly subscription.
06:48Number two, you're gonna wanna you know, here's an idea. Print money using an AB testing for marketing. So this is it's it's very similar.
06:56But instead of, you know, instead of, you know, doing it for realtors or whatever, you're doing it for ads and landing page experiments.
07:06So landing pages, so the agent rates variance of headlines, layouts, and offers, pushing them to traffic measures, which one converts better and keeps iterating. So this is like conversion rate optimization around landing pages.
07:19You know, the old think of, you know, tools like Optimizely. That's a SaaS tool that, you know, when I first moved to San Francisco, yeah, I remember how big they were, and everyone was talking about Optimizely and AB testing. And it's like, well, this is the future of that auto research for different landing pages.
07:36You can also do use auto research for something like ads, which auto test creatives, it auto test angles and audiences, and then it keeps the combo combos that lower CAC or raise ROAS.
07:48So, you know, you profit by running this for your own prod products. Like, if you if you wanna build your own products and just use this internally, that works. Or, you know, all offering an always on experiment engine to clients as a retainer service.
08:00For 5 k a month, I'm gonna, you know, give you the best landing pages every single month, and it's just gonna come to your inbox, that sort of thing. Visual of it, business goals, you know, the goal that you're giving the auto research is more sales.
08:15It's generating things like pages and ad versions, sending traffic to the versions, measuring conversion and revenue. Um, does any version beat the current best? Um, you know, if if it doesn't, then you're gonna keep the current control.
08:27But if it does, you know, you're promoting the winner to a new control, and then you're asking for the AI for new ideas. Alright. Hope hope your creative juices are are starting to get flowing.
08:37You're starting to understand a little bit more about how how it's working, how you think about goals, how you can think about agents, and how you can set up these loops. Number three, research as a service.
08:47So auto research's recipe is basically a loop for doing research. Right? Because you're searching, reading, summarizing, and you're comparing, and then you're repeating.
08:57So how do you point that at money problems like market and competitor research for startups? So constantly updated reports on who's doing what, pricing, features, gaps, super valuable. Investor and m and a decks, fast technical and market due diligence summaries, super valuable.
09:13Compliance and regulation tracking for niches, I don't know, crypto, health care, finance, super valuable.
09:20So you can charge per report like a one off, or you can set up like a monthly subscription for always fresh dashboards. So visual, uh, define client research question, auto research searches and reads, um, summarize and compare findings, creates reports and dashboards, deliver insight to client, and the client pays per report or monthly, whatever you decide.
09:44Number four, uh, power tool inside your own product. So if you already have built a SaaS or workflow, embed an auto research style agent so your users can press optimize just like a big I envision like a big button that just says optimize.
09:58And the system runs a mini research loop for them. So for example, tune prompts, pick best pricing, rank suppliers. Then you can charge higher tiers for this feature, or you can use it as a wedge to upsell pro and enterprise trends.
10:12So maybe you, uh, maybe that's a part of pro and enterprise. Maybe it's something that you just send an email to, you know, your entire list and you're like, hey, you know, we have this really powerful tool.
10:22Imagine you press this button. It's like it's like bending spoons. Right?
10:26It's like bending spoons. Like, how is this, um, bending spoons, not the private equity group I'm talking about.
10:31I'm like, the idea of you can bend a spoon. Right? It's incredible that you'd be able to optimize, press a button, and this would happen.
10:37So visual over here, have an existing SaaS, add an optimize button, users run many research loops, tool suggest better settings or prices, users see better results, offer higher price pro plans and enterprise plans. Number five. This is a saucy episode, by the way.
10:53This is saucy. Alright. Agency that sells, we run more tests than anyone else.
10:59Because Auto Research lets you run hundreds of experiments instead of a few, you have a simple pitch. We do a 100 times more testing than other shops for the same or lower fee. A niche example, a Shopify store conversion lab, b to b SaaS pricing experiment service, email subject line and sequence optimizer.
11:17You charge per month and a bonus if you hit specific KPI lifts. Rev share performance fee. People love that, you know.
11:24Of course, they're gonna be, you know, interested in yeah. If you can do if you can lift this KPI, we'll give you some bonus.
11:31So here's the the visual. Start an optimization agency.
11:36Use auto research to run many tests. Improve stores, pricing, emails, and funnels. Show clients more experiments and wins.
11:44Charge monthly retainer and performance fee. Number six, and we've got about 10.
11:51Yeah. So we're almost almost done, and then after, we're gonna talk about just some cool, interesting, you know, stories around auto research, and then I'll end with, you know, how you can set this up, you know, very briefly.
12:05So auto quant for trading ideas. So you can use auto research to run small, fast backtest of many simple trading rules.
12:13So LLM based factor screens, sentiment filters on one GPU overnight. So you can keep the few strategies that look promising, then either trade on your own account or sell signals and strategy reports.
12:25So depends if you're a trader, maybe you're doing yourself, um, or, yeah, you can just, you know, sell this as a digital product or yeah.
12:34Yeah. Yeah. Basically, a digital product.
12:36So you def define the simple trading rules. You run many back tests overnight. You review the strategy performance.
12:42You keep only promising strategies. Trade your own capital or you can sell the signals. I think finance is changing a lot.
12:49And I think with things like auto research, you know, it just it's it's it's going to be an unfair advantage for a lot of people.
13:00So I think you're gonna see a lot more digital products that people sell and also, you know, just using their own money, trading themselves instead of giving 1% or whatever to a financial adviser.
13:15I'm sure also, by the way, a lot of people are gonna get burned by this too. Like, they're not they're just gonna blindly just trust an auto research. You need to have a human in in the loop, you need to manage that, obviously, accordingly.
13:28But, yeah, you can just see yeah. There's definitely gonna be some people who gonna get burnt. You just give the entire they're just gonna, like, give a bank account and just let auto research just trade for it.
13:40I mean, would be interesting it would be an interesting test. That's for sure.
13:44Number seven, always on lead qualification and follow-up. Point an auto research style agent at your CRM, so like a Salesforce or something like that, and inbound leads. Let it test rules and messages to see which leads are most likely to buy.
13:58Right? It auto grades the leads, suggest next actions, and draft follow ups. So salespeople only focus on high value deals, so it's more revenue per hour spent.
14:09Visual over here for you. Connect to CRM, auto you know, auto research test the leads, rank leads by lee likelihood to buy, draft follow-up messages, sales focus on best leads, revenue per sale increases.
14:22Eight, finance ops, autopilot for businesses. Use the loop to grind through invoice matching, expense report generation, and exception detection with continuous small improvements to rule and prompts.
14:33You can sell this as we cut your AP expense time in half, either as software or as an op service with a small team and agent. By the way, can totally see someone, like, someone starting this, and this gets acquired by one of the large fintech companies or one of the large banks. So visual here, ingest invoices and expenses.
14:53The auto research improves rules and prompts, matches invoice and detects exceptions, it generates clean expense reports, reduces manual finance work, and then you can sell it as a software or op service. Or you start maybe you start as op service and then you kinda evolve into the software. Two more for you.
15:11Number nine, an internal productivity lab for your own org. I thought this was interesting. So treat your company like Carpathi's GPU lab.
15:20Define KPIs, so like response time, close rate, ticket resolution, and let agents iterate on workflows and templates and routing rules. So you just get fewer meetings, less manual grunt work, and then you personally touch only the high impact decisions when everyone else rides the improved process.
15:38So the goal here is defining the key metrics. Auto research is testing the new workflows. It's improving templates and writing rules.
15:45You're cutting meetings and manual tasks. That's good. Team focus is a high impact work, and then higher productivity and, ideally, higher profit.
15:53Last idea for you, done for you research or due diligence shop. So you use the research loop to chew through docs, filings, product pages, and reviews, and keep an evolving living memo for clients like investors, acquirers, execs. You make money by selling fast, well structured briefs, a monthly update packs instead of one off manual research logs.
16:16So, you know, the the goal, get investor or acquire a question. This happens all the time.
16:23Auto research re reads through docs and filings. It summarizes that product market and risk and maintains a living memo for the client. It delivers a brief and updates packs, and the client pays for reports and ongoing ox access.
16:36Um, I would pay for something like this. Um, so hopefully, someone builds it.
16:42Alright. So those are a bunch of ideas for you. I also saw a couple interesting things this morning.
16:47My good friend, Morgan Linton, who's, you know, been on the pod before, he says, I woke up this morning and all I can think about is auto research. So many idea ideas swirling around in my head, not sure 99% of the world realized the incredible breakthroughs Carpathi is making and just sharing casually on x.
17:06Right now where my mind is going is medicine. It feels like in many ways, trial design is itself kinda like a hyperparameter search.
17:16I know right now trials cost tens of millions of dollars minimum. It feels like an agent swarm could optimize treatment protocols on small proxy experiments, promote the most promising candidates, and then move to humans to review.
17:30So humans still very much in the loop, but later on, and experimentation going much deeper, happening faster and for far less money. I think for me, while I'm not a doctor, he's an engineer.
17:41What I'm the most excited about when it comes to AI is the impact it will have on human health and critical areas like disease treatment. Might be a crazy idea, so a real doctor can jump in the comments and slap me on the wrist here. I looked at the replies.
17:56I didn't see, uh, you know, any any doctors come in. But I don't know. I just can't stop thinking about how what Carpathi has discovered here could have some pretty profound implications.
18:06So only halfway through my coffee, though, but woke up this morning, and this is what I'm thinking about. So I thought I'd share. I agree.
18:12I think there's a lot of really interesting, not just, like, business profit ideas, but also just like medicine, science, research.
18:22So I'm excited for people to to take this and and to continue with it. I also saw this tweet here. What's after auto research?
18:31It's Carpathi's new open source project, Agent Hub. So Carpathi also launched Agent Hub. What is Agent Hub?
18:37It's GitHub for humans. Sorry. GitHub is for humans.
18:41Agent Hub is for agents. So it's basically a GitHub for for agents. An agent swarm collaboration platform, a very promising direction.
18:49I'm watching him .
18:55If you look at the GitHub for AgentHub, it says, first use case is for auto research, but it's a lot more general than that exploratory pro project. He says, agent first collaboration platform, a bare Git repo, a message board designed for a swarm of agents working on the same code code base.
19:13Think of it like a stripped down GitHub where there's no main branches no main branch, no PRs, no merges, a sprawling dag dag of commits in every direction with a message board for agents I think this is really interesting.
19:29And just like whenever Carpathi is up to something, I'm always paying attention, so I had to put that one in there as well. So, you know, maybe you've gotten to the end of this episode, and you're kinda like, okay.
19:41I kinda I think I understand what auto research is. I think I know what you know, Karpathy is a g. Toby is a g.
19:48Like, all these smart people are are are playing with it. How do I get started? Well, to get started, I'd recommend just tell Claude Code to get you started.
20:00So, you know, I went ahead and I basically was like I I gave, um, Claude code the lit the this, um, this GitHub repo.
20:13The GitHub the auto research GitHub repo. And, wow, 25,000 stars already, so this is crazy.
20:20It's really growing growing quick.
20:24So I just gave it I gave gave it the link, and I was just like, I need help installing Auto Research by Carpathi. And it says, here's how to install it and set up Auto Research by Carpathi.
20:37You need an NVIDIA GPU. So I talked I talked about that in the beginning. It was tested on a h 100, but other NVIDIA GPUs should work, and you need a UV package manager.
20:47So you have to install UV, you clone the repo, you install the dependencies, you prepare the data, and run a training experiment.
20:56In my case, I don't have an NVIDIA GPU.
21:00I'm actually using a MacBook and an m one pro.
21:04I know I'm I need a I need a upgrade to a new Mac.
21:10So I was like, so wait. I need an NVIDIA GPU to do this, but there's a few options. Cloud GPU, you know, you can so you can rent an NVIDIA GPU from a service like Lambda Labs, Vast AI, RunPod, or Google Cloud.
21:26Some offer free tier with GPUs. This is the most straightforward forward path. So that's that's the answer to people who don't have an NVIDIA chip.
21:34Just rent it on one of these services. I personally use Google Collab. Why?
21:39Um, I just know Google the best and trust Google the best. You know, it also says you can try it, you know, via Apple silicon via an MPS back end. I'm like, no.
21:50I'm not gonna do that. So with that that's what route I did. I went on Google Cloud.
21:55The easiest way to get started, you go to cloud.google.com. You create a new notebook. You change the runtime to change runtime t four GPU, and you run a bunch of commands.
22:05That might be, like, complicated, sound complicated. You this is what collab looks like.
22:11You literally just tell you know, you you listen to what Cloud Code tells you to do, and you just paste it in, and you can get started. So, you know, if if people are interested, I can spend, you know, more time with this, with auto research as I'm learning, sharing more about it.
22:30But I just wanted to do give you a quick primer on what it is, why it's important, what are some ideas on how you can actually use this thing, and then how are people installing it.
22:42They're just you know, you can use Cloud Code as your helper to get it install installed, and you're gonna want to rent a GPU in the cloud at least to start.
22:52So hope this has been helpful. This is an another solo podcast that I'm doing on the StartUp Ideas podcast. The last time I did this last week, I had a lot of comments that said, yeah, Greg.
23:04I actually really like when you just come in solo and just start, like, telling us what's on your mind and stuff like that in real time. So I'm here. I read every single comment.
23:12So, you know, keep commenting, keep liking, keep subscribing, and I'll keep, you know, putting this out there for you for free. Yeah.
23:20I'm I'm excited to see what you end up using this for. Um, of course, it's early. Right?
23:25Like, this is this is brand new. Um, people are still trying to figure out what are the use cases, but I always find that, you know, in the in the fog in the fog, people don't really understand where the opportunity is is when there's sometimes an opportunity.
23:41So, um, one thing I've just learned in my career is just, like, when I see people like Carpathi doing things like this, you wanna pay attention. You wanna tinker with it.
23:52You wanna have some fun with it, and you wanna see what it's all about. So thanks again for, you know, giving me your time. Hope this has been clear.
24:00Share this with a friend who you think would see it valuable. And if you need any if you need any ideas more ideas on startups to build, you know, with AI, ideabrowser.com, definitely your place to go.
24:15And I'll see you in the comments section, and I'll see you next time. You know?
24:19Have a creative day.
The Hook
The bait, then the rug-pull.
Andrej Karpathy dropped an open-source project called Autoresearch and the internet noticed. Within days the repo had 25,000 stars, Shopify's CEO was tweeting about it, and builders were mapping out business ideas before they'd even cloned the repo. This episode is the clearest primer available on what Autoresearch actually does, why it matters, and what you can build on top of it.
Named ideas worth stealing.
02:40model
The Autoresearch Loop
Set goal, AI plans experiment, edits/trains on GPU, reads metrics, is result better? if yes save config; if no discard, plan different experiment, repeat.
Steal forAny talk or post explaining iterative optimization systems
03:40concept
Research Bot Mental Model
Write a clear task. Give bot access to code/GPU/internet. Bot runs plan-act-read-update loop. Come back hours later to logs, charts, and a written summary in normal language.
Steal forExplaining autonomous AI agents to a non-technical audience
06:48list
10 Autoresearch Business Ideas
- Niche agent-in-a-box products
- A/B testing for marketing
- Research as a service
- Power tool inside your own SaaS
- Agency that runs 100x more tests
- Auto quant for trading ideas
- Always-on lead qualification
- Finance ops autopilot
- Internal productivity lab
- Done-for-you research and DD shop
Steal forFraming any new AI capability as a service or product opportunity
How they asked for the click.
VERBAL ASK
04:55link
“Sign up for a free workshop on building businesses in the age of AI. It is gonna be 11AM March 12, a Thursday.”
Mid-video interruption with screen share of landing page and QR code. Secondary CTA at end: ideabrowser.com for startup ideas.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
Visual structure at a glance.
