Brett Adcock

We have two additions to PersonalAgentBench: @hark_labs and @OpenAI's Dots! Dots has taken the lead, with Hark close behind while Gemini Spark moves to third, followed by Instinct, Grok Bot, and Muse. We also doubled the number of tasks this week, adding research, image understanding, editing, and everyday communication. The wider set changed the picture. Gemini Spark was strongest on research but weaker on image and editing tasks. Hark and Gemini tied on tasks involving user authority. Dots was the most consistent across task types, although it was weaker when it had to infer personal context. No assistant led every category; the overall ranking rewarded consistent performance across the full set. This is still an early snapshot, and we’re excited to see how results evolve as we add more agents, expand the tasks, and re-evaluate after product updates. If you’re using any of these six personal agents, follow along by visiting the link in the original post.

How cool is this. @WHOOP widget in my Hark dashboard.

@mustafa_2vec @prit4k @hark_labs I gave Hark and Muse the exact same prompt today. I wanted to check availability for a massage and book if available. Muse told me the facility doesn’t take online bookings (false). Hark figured out availability and booked the massage.

Pretty impressed by Hark @hark_labs after just a day I’ve tried Muse, Instinct, Fo, Grok Bot, and Dots, but Hark is the first assistant that’s actually been deeply proactive. For example, it noticed I’d booked an Airbnb for a weekend trip in April 2027 about 3 months ago, but hadn’t booked a flight, then asked if I wanted help finding one It feels like Hark goes deeper into your data and actually connects the dots instead of just waiting for you to ask it something. The UI is great too, my only wish is native liquid glass on iOS!

Hark has exploded over the last 48 hours, and all the user growth caused some infra issues today. fwiw launch was far beyond my expectations I'm personally on it with the team; we're actively working to stabilize

We designed Hark Pro to think a step ahead, live in one simple thread, and spin up custom interfaces for your needs. It lets you offload tasks with the press of a button. We set out to make an experience that works on your behalf approachable to everyone. Proud to be working with the crew at Hark every day. t.co/ARloOOCL4r

Just found @adcock_brett through @hark_labs . It’s his AI company, and Hark is just way better than OpenAI’s Dots. And that’s not even all of it. He runs 4 companies at once. • @figure_robot — humanoid robots • @hark_labs — AI • @cover_thz — weapon detection • @flyArcher — flying cars I have no idea why he isn’t discussed as much as Elon.

This is an incredible full circle moment. Brett and I went to Vietnam in 2025 for a manufacturing trip, and the process of filling out this visa form honestly took so incredibly long. It was actually a task that we promised ourselves that we needed to be able to do. To see that Hark is able to handle this for someone brings me so much joy. It’s wonderful to walk into this future where we can focus on what’s important. Thanks for sharing Jason!

Since this post, I’ve been blown away by Hark’s ability to find relevant sources in my life with no context and tackle them. What?! x.com/mysteryethos/s…

This new AI agent blew my mind Hark is a proactive agent with a custom interface that completely transforms based on who you are They gave me early access a week ago and I've really enjoyed using it Here's everything you need to know about Hark:

Hark is an awesome AI agent Was lucky enough to have early access over the last week Some things I really like about it: • Best interface in agents. Period. Beautiful to look at and also completely customizable • Extremely proactive. Constantly giving you ideas on what it can do for you • Really strong browser use. Has it's own cloud computer that it gets tasks done in lightning fast Still waiting on computer use so it can manage my computers, but I'm sure that's coming soon Really enjoyed using it. Highly recommend trying it out

Wow, @hark_labs made a straight up superior @Muse. Its harness fully supports computer use (pretty fast, at that) in an isolated sandbox with vault-based auth you can reuse across any app (huge for Google), and you can actually see the "computer" being used in the cloud. Plus, its UI is so clean and such a refreshing take compared to most cluttered 3-column chatbot interfaces. Easily the best personal AI offering on the market right now and the most "task-capable" agent on the AI growth curve.

Hark launches this week. The first 100,000 sign-ups get a paid plan for free I don't do anything without Hark anymore. The team has obsessed over every detail, and it shows. Can't wait to see how you all use it Join the waitlist: hark.com

100,000 free paid plans. That is how Brett Adcock is launching Hark this week. The founder @adcock_brett announced on X: the first 100,000 sign-ups get the paid plan free. Not free for a while. A hard cap. That turns a launch announcement into a race, and everyone who misses the cap becomes the next wave of demand. The product backs the pitch. Hark Handoff scores 97.7 on the Online-Mind2Web web-agent benchmark, ahead of GPT 5.4 at 92.8 and Claude Opus 4.8 at 84.1. $700M raised at a $6B valuation, and the whole launch is one post with a number in it. Steal this: if your launch offer has no cap and no deadline, it is not an offer. It is a pricing page.

.@adcock_brett explains why @Figure_robot pivoted away from designing their humanoid robot's hands after human hands: "This is extremely important. If we want to solve AGI and get to human intelligence in the physical world—it's all going to start with the hands." "Our first-generation hand is a tendon-driven hand." "The rationale here is that a human hand is like this: most of our motors are in our forearm and little tendons drive all the fingers." "It ended up becoming the wrong engineering choice." "We recently unveiled our high degree of freedom, next-generation hand a month ago. This hand has human-level dexterity, as many joints in the hand as a human."

F.02 Decommission

Hasta la vista, F.02. x.com/figure_robot/s…

Goodbye F.02

Today, with over 100 industry partners, we introduced the NVIDIA Open Agent Safety Platform, bringing together OpenShell and Sentry. Artificial intelligence is extraordinary technology that will advance discovery, productivity, security, health, and prosperity for generations to come. But its full promise can only be realized when people have confidence that AI is being built to be safe and deployed with wisdom and responsibility. This is bigger than a single product. It's the beginning of an open ecosystem to build the trust layer for safe agent systems. Together, we are building the foundation of the AI economy. Trust and innovation are not in conflict. Safety is how trust is earned. We must build not only the most capable AI, but the most trusted AI, so that this extraordinary technology can realize its enormous promise for the world. t.co/ugYWQ1MyRi

Watch 4 straight hours of Helix 2.5 at work in 30 homes, with no additional training.

I always thought humanoid robots look like us because the world is built for humans (doors, stairs, kitchen counters etc) but figure's new release shows another reason, and i think it's the bigger one: > until now, teaching a robot a chore meant collecting data on that exact robot in that exact room, which is slow and expensive and only ever taught it that one thing > figure built an app where people get paid to film themselves doing everyday stuff from their own POV, and about 90,000 people a week are uploading to it now > the robot learned how humans move and handle things from those videos, and then got a bit of extra training on three specific jobs > then they put it in 30 homes it had never seen, with towels, beds and toys it had never seen either, and no training happened in any of those houses > it made the beds, folded the towels and tidied the toys into a basket > the same robot without the human videos finished the job 9% of the time, and with them it finished 56% of the time > and the more human video they feed it, the better it gets at guessing the next move, in a curve you can predict (that's key, it scales!) this only works because the robot has our body. a robot on wheels with a claw can watch you make a bed all day and copy nothing, because it doesn't have the hands or the reach so the human shape also solves the problem robots always had, which is that nobody ever filmed enough robots doing chores. there are 8 billion of us doing them every day though, and now filming us is the training it still fails almost half the time, so nobody is getting a housekeeper this year. but going from 9 to 56 in one step is a big jump the flywheel is flywheeling

LIVE Interview With Figure CEO | AI Breakthrough, Helix 2.5, Zero-Shot Generalization and More x.com/i/broadcasts/1…

Today may the pivot point where I start taking @Figure_robot seriously.

We just visited Figure and it was 94 degrees out and the robots were everywhere outside and had no issue with the heat These robots are built to last

This is the part of humanoid robotics that gets interesting fast. @Figure_robot already has multiple robots moving around its San Jose HQ without someone sitting there controlling every step. The goal isn’t just making a robot perform a demo. It’s getting to the point where robots can operate on their own in the real world.

Gotta love @Figure_robot At its San Jose headquarters it has a bunch of robots walking around with no nerds watching them.

Introducing Index Today we're coming out of stealth with Index, the largest & most diverse robot dataset in the world → 30min of video uploads/sec → 16M video uploads → Paid $15M to date → 264k downloads We're committed to spending $1B the next 12 months on data & compute

@Figure_robot just stacked the two halves of the Physical AI bottleneck a week apart. First Index: a creator network collecting real-world task video at roughly half an hour of uploads every second, across homes and workplaces in 100+ countries. That is the long-tail data you do not get from a lab teleop setup. t.co/kW0lsLgK0l Then Nscale: $3.5B of compute to start, intent past $6B, up to 100,000 NVIDIA Vera Rubin GPUs. t.co/HNvkct0Vbf @adcock_brett said they are largely constrained by data and compute. Index is the data bet. Nscale is the compute bet. Figure's own writeup says the quiet part: data alone cannot solve this. My take: Compute is a prerequisite. Collecting more data is critical. But finding the 1% of data that actually teaches the model is what compounds. Search and curation, not just capture. That's the problem we are focused on solving at @Foxglove.

buying a house on mortgage in San Francisco x.com/adcock_brett/s…

Exciting day for NVIDIA and @huggingface. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. They allow every developer, startup, university, industry and country to build with, customize and benefit from AI. Thank you @ClementDelangue for coming to me. NVIDIA is going to be a great home for Hugging Face, its community and the future of open models. 🤗 t.co/q8Om2Xc5ye

Putting an intelligent, general purpose robot in every home will require massive scale in both data and compute. Today, Figure announced a $3.5 billion deal to scale Helix 🧬 to up to 100,000 Vera Rubin GPUs. There’s never been a better time to join Figure. x.com/adcock_brett/s…

brett adcock is the epitome of serial entrepreneurship in this decade vettery → sold for $110M to adecco archer aviation → evtol aircrafts,~$2.7B IPO figure AI → humanoid robots, $39B cover → weapon-detection AI for schools hark → personal intelligence, 2025 index → largest robot dataset in the world 🚀

Four years ago, @Figure_robot didn’t even have a robot. Today, it’s valued at $39B, has built 1,000+ humanoids, deployed robots at BMW, built its own AI models, manufacturing infrastructure and a global data collection network. What impresses me most is the speed. Figure isn’t building just a robot. It’s building the entire stack at the same time: hardware, AI, data, factories and real-world deployments. And now Figure 04 is already on the way. Four years later, much of the humanoid industry is being measured against what @adcock_brett and the Figure team have built.

Introducing Index Today we're coming out of stealth with Index, the largest & most diverse robot dataset in the world → 30min of video uploads/sec → 16M video uploads → Paid $15M to date → 264k downloads We're committed to spending $1B the next 12 months on data & compute

This was few months ago. I wonder how capable they are now 👀🤖 @figureonsol @Figure_robot @adcock_brett x.com/CyberRobooo/st…

Handoff knows when it's a have-a-coffee-waiting kind of Monday.

This is hard to achieve, because it must continuously coordinate vision, grasping/hand placement, leg stepping, weight shifting, & balance. Any mismatch risks a fall from height. This couples perception, planning, + dynamic control in a continuous task—harder than flat-ground walking or isolated manipulation. I think Figure relies on its Helix system, upgrades incorporate real-time visual input alongside proprioception, enabling the robot to adapt w AI approach rather than follow purely pre-programmed controls. ~15 ft both up and down on a taller structure increases the challenge (descent requires careful load management and can be less forgiving). @Figure_robot is making some real progress. 👍👍🚀

You may or may not see Figure robots walking outdoors when driving around the Figure headquarters. Just saying. Saw some yesterday with @GoingBallistic5

If you visit Figure’s HQ, you’ll see their humanoid robots wandering around the office.

Today, we’re partnering with AT&T to bring you the next generation of AI-native consumer devices. AT&T is also investing in Hark and collaborating in areas like device certification and network connectivity. AT&T’s infrastructure paired with Hark’s AI platform and hardware will help enable Hark to become the leader in personalized consumer intelligence.

@adcock_brett You should melt them.

Good summary of the major points of our robotics investment thesis Worth understanding as the industry continues to accelerate x.com/intern/status/…

.@adcock_brett. Founder of Figure, Hark, and Archer. New episode is up. @thesamparr @ShaanVP

.@adcock_brett explains why @Figure_robot pivoted away from designing their humanoid robot's hands after human hands: "This is extremely important. If we want to solve AGI and get to human intelligence in the physical world—it's all going to start with the hands." "Our first-generation hand is a tendon-driven hand." "The rationale here is that a human hand is like this: most of our motors are in our forearm and little tendons drive all the fingers." "It ended up becoming the wrong engineering choice." "We recently unveiled our high degree of freedom, next-generation hand a month ago. This hand has human-level dexterity, as many joints in the hand as a human."