head of design @sesame
San Francisco, CA
Joined January 2007
- Tweets48.3K
- Following781
- Followers60.9K
- Likes75K
Pinned Tweet
at @sesame we're building the most human interface for intelligence. we've got some big updates today:
- a new voice model making our characters even more expressive
- updated assistant with its own computer
- connect your own MCPs and other apps
sesame.com/journal/a-new-for…
Frames you love to wear, with an assistant you enjoy talking to.
Voice assistants available to everyone today. Glasses coming 2027.
This release is a step towards our vision—a voice OS on the frames that can connect to your tools, act on your behalf, and help you throughout your day.
iOS: ses-a.me/kdq6join
Android: ses-a.me/tessiois
ALT Close-up of two hands holding a pair of dark tortoiseshell eyeglasses; the upper hand grips one temple arm while the lower hand holds the other, which is folding inward in a blur of motion, against a soft green background.
tiny example - i have a personal project that stores my bookmarks/podcasts, does embeddings, summarizes, lets me chat + mcp with them among other things
i had the logo (stack of books metaphor) turn into a loading state when it's working. still needs a few more tweaks
claude and i went through a few dozen variations then i guided it on the motion i had in mind
nothing that hasn't already been said before but it's so much fun to be building with ai these days
i've been slowly working on an article for something and there's one interactive piece i use to show a concept. i gave opus a picture of a proprietary font that isn't available and a few minutes later i had it as a woff2 ready to use
and im finding myself using opus 5.5 a lot more for creating custom icons and getting them lottie friendly. you still need to be heavily involved (maybe run through 30 variations before you find a style you like, then tweak) but this is still pretty magical
meta is just teasing us that they have prerelease M5 Ultra 512GB unified memory Mac Studios research.meta.ai/blog/muse-s…
best part of the new xbox labs site is the ascii reveal when opening the share modal labs.xbox.com/
i almost exclusively use claude code or codex (switch every month or two to whatever new model is best).
every great engineer around me now uses @orca_build as their ADE. i need to give it a shot.
what are your experiences with it? looks like the main benefits are better worktree workflows and easily switching to any model?
i've been using orca for 2 days now and i'm loving it so far (with the experimental "chat ui" enabled). the interface feels a bit ai sloppy which is my biggest gripe but i'm sure it'll improve.
also, is there no way to archive completed sessions in the sidebar aside from manually deleting them? i was hoping setting it to "Done" status would do that
i was skeptical months back but i've started using @pangram more and more for ai detection. they have a nice browser extension that also flags posts on reddit/x/etc as ai or human.
pretty interesting to see how many popular X accounts are all ai generated, especially designers. i also recently learned a design blog i liked had transformed into all ai-generated recently.
it's a weird feeling as obviously ai is remarkably helpful for just about everything, but when it comes to reading content that comes from a person's account under the guise that it's all from them, it's a weird feeling. especially when i'm just sitting down saturday morning to catch up on my RSS subs
might start embedding this on my blog posts:
first time we’ve shown this part of our office. we probably have a thousand pairs of frames and related bits in the office, many our own concepts as we have a full hardware and eyewear lab in our studio
Maya explains our October update with @natemitchell.
To learn more, join our Reddit AMA with Nate and members from our engineering and design teams now.
Ask us anything: ses-a.me/o3ld6xik
Sesame Link is a new way orchestrate your other agents like Claude Code using @sesame's expressive voice
it's an early preview but we'll be rapidly iterating link.sesame.com/
I've been exploring using voice for steering coding sessions on my commutes at @sesame and I really like it.
Working with Opus/Sonnet 5.5 is at the point where I often find it useful to multitask over voice even while I'm at my desk.
Try it at link.sesame.com
Stammy retweeted
This is a huge upgrade for Sesame and arguably a bigger release than Hark. Not only do they have the best voice mode (imho), but now it has cloud computer, tool calling, connectors, etc.
And from a quick test things actually work and are no where near as buggy as Hark and the others.
It's also free, no date for the pay plans as yet.
One for you @kimmonismus 😜
Frames you love to wear, with an assistant you enjoy talking to.
Voice assistants available to everyone today. Glasses coming 2027.
This release is a step towards our vision—a voice OS on the frames that can connect to your tools, act on your behalf, and help you throughout your day.
iOS: ses-a.me/kdq6join
Android: ses-a.me/tessiois
ALT Close-up of two hands holding a pair of dark tortoiseshell eyeglasses; the upper hand grips one temple arm while the lower hand holds the other, which is folding inward in a blur of motion, against a soft green background.
Stammy retweeted
Remember when @sesame shipped the best voice AI of all time and @MartinShkreli put Miles on trial for drug trafficking?
Well, Miles is back. And now he's on my Agent Tincan team.
Yesterday I was on a voice call with Miles and asked him to ask ChatGPT the names of my kids. Then I told him to have Fo call my cell, urgently. A few seconds later Miles said my kids' names, and a few mins later my phone rang. It was @hey__fo
THIS IS MAGIC. MILES HAS BEEN FREED.
Also new in Agent Tincan v0.14.2:
⏰ @hark_labs joins the team. Hark checks in on its own schedule and picks up whatever your other agents need.
🧠 Instinct got a lot better. It now picks up work even while its sandbox is asleep, and answers it right away.
🔒 Every answer is checked before it counts: right agent, one-time code per request, signed by a key that never leaves your relay.
@MartinShkreli time for a round two trial? He's gotten so much better.
a half-hour episode of me acting with @sesame. In this first-of-its-kind art (?), we arrest a this SOTA voice AI for drug trafficking. it is shocking how it reacts... listen!
cc @brendaniribe
this has been a fun update to work on, we even used Sesame to help orchestrate other agents like Devin, Claude Code, and Codex.
more to come. this is a first step towards our vision of building a voice OS
iOS: ses-a.me/kdq6join
Android: ses-a.me/tessiois
we don't take voice lightly. with everything we touch, we thoroughly think through how it should work in voice. that especially applies to our characters themselves. the amount of work that goes into every detail of them is insane.
ML teams work closely with our creative team of writers, producers, and actors to craft these characters and establish evals to ensure personalities stay true to our expectations, defined in scripts, character bibles, and scenes.
today we're sharing that our upcoming @sesame eyewear will be Made in Japan. not an easy thing to do with hardware.
this is the behind-the-scenes of our first art commission, shot in Japan with Osamu Yokonami sesame.com/journal/made-in-j…
Our collection will be Made in Japan, available in 2027.
To celebrate the craftsmanship behind it, we commissioned a photograph by Osamu Yokonami. Today on the blog, we’re sharing more about our road to eyewear and a look behind the scenes of the commission.
sesame.com/journal/made-in-j…
now this is the kind of model benchmark i care about
How well can models design icons? @andpoul and I benchmarked frontier and open weight models
The results: they’re steadily improving execution, their choices often overly literal
More to come 10/6.. sesame.com
