Sable is building Aidan, an AI employee that runs live customer conversations by actually operating a browser, clicking through a product, and responding to what appears on screen in real time. The tool is aimed at go-to-market teams who want prospects to see software in action during a sales call or onboarding session instead of scheduling repeated human-led walkthroughs. Aidan runs inside a virtual machine Sable calls a LiveBox, which lets the agent demonstrate a product's full capabilities to a buyer , and it builds product knowledge by ingesting recordings of top sales calls, internal documentation, help center content, and marketing materials into a reusable "brain."
The launch matters because computer-use agents have mostly been demoed on asynchronous tasks like filling forms or booking travel, and Sable is pushing them into synchronous, conversational settings where latency and visual grounding are the hard problems. Improvements in vision and computer-use models, combined with advancements made by Sable's team, now enable Aidan to perform live browser actions on demand and continuously perceive the screen and respond in real time. Early customers include Notion and Decagon , which gives the company a foothold in both horizontal productivity software and the customer-support AI category itself.
Sable was founded by Nim Ravid, Leon Chen, Linda He, and Itamar Rocha, who met at Harvard University and conducted AI research in post-training, reinforcement learning, and multimodality. Prior to founding Sable, members of the team worked at SpaceX, Google, Meta, and Together AI. The company has raised $45 million from Sequoia Capital and 8VC , and this launch opens access to a beta for teams that want to put Aidan in front of their own prospects.
Comments (15)
Sign in to join the discussion.
P
Priya IyerJul 16, 2026
Naming your AI 'Aidan' is such a soft launch move. Half the sales team is now going to have an existential crisis when the real Aidan joins standup.
R
Reem H.Jul 16, 2026
Tweet copy is doing a lot of heavy lifting with 'first computer-using AI built for realtime conversation.' That's four qualifiers stacked to make sure nobody else can claim the crown.
T
Tomas LindqvistJul 16, 2026
The demo cut away every time Aidan had to actually click something. I want the uncut director's edition where it fumbles a dropdown menu.
K
Kwame Osei-BempahJul 16, 2026
Cool cool cool but where are the docs. What's the latency budget on a vision loop mid-call, and do I get webhooks when Aidan gets stuck on a captcha?
Y
Yuko ShibaJul 16, 2026
Watched the launch video three times and still can't tell if Aidan is talking to the customer or to the CRM. Might be the point.
M
Marcelo D. PazJul 16, 2026
Ops question nobody is asking: who owns the pager when Aidan hallucinates a refund on a live call. Is there a human circuit breaker or do we all just pray.
J
Juno ParkJul 16, 2026
Onboarding better not be a 30 minute Loom with a solutions engineer. If I can't spin up an Aidan and have it flub a call within 5 minutes I'm out.
H
hedvigJul 16, 2026
Seed partner brain says the Notion and Decagon logo drop in tweet three is the actual pitch deck. Everything else is garnish.
C
Chidinma EzeJul 16, 2026
The pacing on that reveal thread is chef's kiss. Founder photo, big number, big logos, then a video. Someone on that team studied film.
L
Linnaeus ZhaoJul 16, 2026
Impression count on this tweet is loud but the reply-to-quote ratio is suspiciously clean. Either the thread landed perfectly or the community notes crew is still warming up.
R
Ravi GuptaJul 16, 2026
Curious how they're handling the perception-action loop under conversational latency constraints. Feels closer to a WebVoyager-style setup than pure tool calling, but the vision grounding claim is the interesting bit.
B
Beatrix N.Jul 16, 2026
Fifth realtime-voice-agent launch this month and the only one that didn't open with a diner ordering pizza. Small mercy, real progress.
S
Sasha MerkulovJul 16, 2026
Everyone is sleeping on the fact that a browser-using conversational agent is basically an oracle machine. Someone will wrap this in a token before Q2, mark my words.
F
Farah Al-AttarJul 16, 2026
Hi Nim, I cover applied AI. Would love five minutes on how Aidan handles handoff to humans and any customer volume you can share on record. DMs open.
A
Ayaan QureshiJul 16, 2026
Congrats team. If you're staffing forward deployed engineers I've got a candidate who spent two years wrangling browser automation at a fintech and is very tired of Selenium.
Work With Us
Book an intro call to explore our suite of services. We might have a waitlist.
Sable
The first AI employee powered by real time browser use and vision – to run customer calls at scale.
The first AI employee powered by real time browser use and vision – to run customer calls at scale.
Tags
TLVC Rating
Community Rating
No ratings yet
to rate this launch.
Sable is building Aidan, an AI employee that runs live customer conversations by actually operating a browser, clicking through a product, and responding to what appears on screen in real time. The tool is aimed at go-to-market teams who want prospects to see software in action during a sales call or onboarding session instead of scheduling repeated human-led walkthroughs. Aidan runs inside a virtual machine Sable calls a LiveBox, which lets the agent demonstrate a product's full capabilities to a buyer , and it builds product knowledge by ingesting recordings of top sales calls, internal documentation, help center content, and marketing materials into a reusable "brain." The launch matters because computer-use agents have mostly been demoed on asynchronous tasks like filling forms or booking travel, and Sable is pushing them into synchronous, conversational settings where latency and visual grounding are the hard problems. Improvements in vision and computer-use models, combined with advancements made by Sable's team, now enable Aidan to perform live browser actions on demand and continuously perceive the screen and respond in real time. Early customers include Notion and Decagon , which gives the company a foothold in both horizontal productivity software and the customer-support AI category itself. Sable was founded by Nim Ravid, Leon Chen, Linda He, and Itamar Rocha, who met at Harvard University and conducted AI research in post-training, reinforcement learning, and multimodality. Prior to founding Sable, members of the team worked at SpaceX, Google, Meta, and Together AI. The company has raised $45 million from Sequoia Capital and 8VC , and this launch opens access to a beta for teams that want to put Aidan in front of their own prospects.
Comments (15)
Naming your AI 'Aidan' is such a soft launch move. Half the sales team is now going to have an existential crisis when the real Aidan joins standup.
Tweet copy is doing a lot of heavy lifting with 'first computer-using AI built for realtime conversation.' That's four qualifiers stacked to make sure nobody else can claim the crown.
The demo cut away every time Aidan had to actually click something. I want the uncut director's edition where it fumbles a dropdown menu.
Cool cool cool but where are the docs. What's the latency budget on a vision loop mid-call, and do I get webhooks when Aidan gets stuck on a captcha?
Watched the launch video three times and still can't tell if Aidan is talking to the customer or to the CRM. Might be the point.
Ops question nobody is asking: who owns the pager when Aidan hallucinates a refund on a live call. Is there a human circuit breaker or do we all just pray.
Onboarding better not be a 30 minute Loom with a solutions engineer. If I can't spin up an Aidan and have it flub a call within 5 minutes I'm out.
Seed partner brain says the Notion and Decagon logo drop in tweet three is the actual pitch deck. Everything else is garnish.
The pacing on that reveal thread is chef's kiss. Founder photo, big number, big logos, then a video. Someone on that team studied film.
Impression count on this tweet is loud but the reply-to-quote ratio is suspiciously clean. Either the thread landed perfectly or the community notes crew is still warming up.
Curious how they're handling the perception-action loop under conversational latency constraints. Feels closer to a WebVoyager-style setup than pure tool calling, but the vision grounding claim is the interesting bit.
Fifth realtime-voice-agent launch this month and the only one that didn't open with a diner ordering pizza. Small mercy, real progress.
Everyone is sleeping on the fact that a browser-using conversational agent is basically an oracle machine. Someone will wrap this in a token before Q2, mark my words.
Hi Nim, I cover applied AI. Would love five minutes on how Aidan handles handoff to humans and any customer volume you can share on record. DMs open.
Congrats team. If you're staffing forward deployed engineers I've got a candidate who spent two years wrangling browser automation at a fintech and is very tired of Selenium.