ego

neolab making ai fun again through human-like voice, memory, intent(@ycombinator W24)

[ego @egolabsai]

@AISinEater Sign up here https://t.co/KgFGIS0O6f

[ego @egolabsai]

@humbldesign @HenryAI_cre @OpenCall_AI @inworld_ai Thank you!

[ego @egolabsai]

@sav_theo @humbldesign @HenryAI_cre @OpenCall_AI @inworld_ai We're proud of our artists here at ego

[𓏴𓀿𓁼𓀐𓏴 @404vish]

police bodycam footage from exactly 1 year ago when i was 15 mins away from death in the mission, san francisco over a random, unprovoked attack. i'm happy to share that i've made an almost complete recovery and have returned to leading @ego_ai_ in spite of multiple complications that landed me in the hospital in tokyo several times in 2025 as a direct result of this attack. thank you to the @ycombinator community & @garrytan for your support thru the worst year of my life & thank you to @scootykins for helping me edit this footage. i sincerely hope san francisco can fix this—keep us canadians safe on your streets please.

[ego @egolabsai]

Artwork by https://t.co/4M6dBVoiCN

Discord link here!
https://t.co/UokFqunfWX

[ego @egolabsai]

The best holiday parties are digital.

We’re opening up the ego AI Discord for the season. Hop in to chat with the team about our upcoming projects, share some feedback, and play The Accomplice

Come for the game, stay for the alpha on what we’re building. See you in there.

[ego @egolabsai]

good morning what are we building today

[ego @egolabsai]

@monali_dambre Congrats!

[ego @egolabsai]

Synthetic voice data is a vanity metric. If you evaluate on clean, generated speech, you are optimizing for a feedback loop, not reality. We are shifting to fully authentic acoustic benchmarks. If the model can't handle the mess of a real microphone, it’s not a product.

https://t.co/6FZPg9riKk

[ego @egolabsai]

Stop training on synthetic speech. You are optimizing for a robot's ear, not a human's. VCB Bench (Oct 2025) argues for 'authentic' audio evaluation to solve the instruction-following gap.

We are currently sourcing research on Human AI interaction, reply with your favorite!

[ego @egolabsai]

@OfficialLoganK he iterated in a cave, with a box of scraps

[ego @egolabsai]

under 5'10 but my context window is 8k tokens 😳

[alli @sonofalli]

guy who’s 5’7 but works in growth

[ego @egolabsai]

@martics_ This is great, we're using this and adding to our discussion on ai and multi-user agent conversations. Perhaps to make a larger discussion feel more natual?

[ego @egolabsai]

Voice AI is failing at the social layer. Latency is solvable, but handling interruptions and awkward silences is an architectural problem. We are compiling a master list of research on human-agent interaction to solve this. Reply with the one paper that changed how you view voice UX.

[ego @egolabsai]

@SofiaMarin5555

[ego @egolabsai]

This paper benchmarks how models handle barge-ins and backchanneling. If your agent can't handle an interruption, it’s just a radio.

https://t.co/zMmk4bhSFm

[ego @egolabsai]

Our second paper on this topic would be Talking Turns: Benchmarking Audio Foundation Models on Turn-Taking Dynamics. Most voice agents fail because they don't know when to shut up. This paper proves that "silence modeling" is just as important as token generation.

[ego @egolabsai]

Voice AI is failing at the social layer. Latency is solvable, but handling interruptions and awkward silences is an architectural problem. We are compiling a master list of research on human-agent interaction to solve this. Reply with the one paper that changed how you view voice UX.

[ego @egolabsai]

Current benchmark: Albert et al. on why passing as human is less important than passing as socially competent.
https://t.co/Enb2iu7Fm0

[ego @egolabsai]

@NousResearch Super exciting! Anything to compete with the big boys.

[ego @egolabsai]

Dialogues with Leaders: Networking with Experts in Human-Centred Design

Time: Friday, 12 December 2025, 3:00 pm – 4:30 pm

Venue: SUTD Lecture Theatre 4 (Building 2, Level 4)

https://t.co/i62Uj33ZMY

[ego @egolabsai]

Next week, our CEO Vish (@AISinEater) takes the stage at Singapore University of Technology and Design (SUTD) to discuss the future of Human-Centred Design.

We’ll be showcasing how ego is moving beyond simple chatbots to create believable AI characters.

[ego @egolabsai]

He’ll be breaking down our whitepaper "Behavior is All You Need" and sharing how we secured backing from various industry leaders.

We are also looking for 2 AI Researchers in SG. If you're ready to ship code that changes the world, you don't want to miss this.

Details + RSVP👇