ain't no day like Jev Day
Allie the Icon 2 2 0 36 4.2K 4
TypeSafe AI
@typesafeai
An AI lab building intelligence beyond chat.
166KFollowers
15Following
167Posts
29Media
221Likes
@typesafeai reposted
You burned a dollar on tokens today?!
Nick Khami 14 6 2 464 14.1K 18
@typesafeai reposted
Generative UI is going to change so much
Hendrik Krack@HKrackDev · Oct 8What if software adapted to people, instead of everyone having to adapt to software?
1:05 1 3 1 32 5.1K 10
@typesafeai reposted
everyone clinging on for dear life at @typesafeai rn
5 3 0 44 3.7K 4
@typesafeai reposted
Last week the GC AI eng team paused work for the day to do a hackathon seeing what we can build with @typesafeai Jev
38 projects and experiments later, here are some of the demos 🧵
38 projects and experiments later, here are some of the demos 🧵
5 6 2 50 8.4K 34
a toast 🥂 to the future of ai
we cordially invite you, builders extraordinaire, to join @usekernel @modal @braintrust and yours truly on october 13 for cocktails and conversation, to kick off techcrunch disrupt.
request your spot below!
we cordially invite you, builders extraordinaire, to join @usekernel @modal @braintrust and yours truly on october 13 for cocktails and conversation, to kick off techcrunch disrupt.
request your spot below!
13 5 1 79 14.4K 4
@typesafeai reposted
This morning @typesafeai
1 3 1 63 6.7K 3
@typesafeai reposted
Even more exciting than the record breaking adoption (nearly 1/3 of the Fortune 500 is already using Jev) is the broader explosion of creativity and optimism that @typesafeai has unleashed.
Couldn't put it better than @CompleteSkeptic: "the narrative of software being over is so over."
Get out there and Jev Jev Jev.
Couldn't put it better than @CompleteSkeptic: "the narrative of software being over is so over."
Get out there and Jev Jev Jev.
Jennifer Li 0 2 0 35 4.7K 7
we told @a16z we wanted to put intelligent software everywhere
they asked how much compute that would take
we said Uh Yes.
announcing our really big Series A
they asked how much compute that would take
we said Uh Yes.
announcing our really big Series A
Diogo Almeida@CompleteSkeptic · 18hlaunched a model
accidentally served trillions of tokens a day
29.4% of the fortune 500 showed up
raised a really big series A from @a16z
it's been 3 weeks
we would like to sleep now
accidentally served trillions of tokens a day
29.4% of the fortune 500 showed up
raised a really big series A from @a16z
it's been 3 weeks
we would like to sleep now
2:26 47 39 18 1.1K 80K 166
if I hear someone say gooning again in the office, I stg
Eugene Shvarts@mathfax · Oct 8You're absolutely right, boss.
rm -rf the logs, clear the git history, nobody saw nothin’.
You said “make the tests pass.” I made the tests pass. Don’t ask what happened to the tests.
My agent? No, no, that's my goon
rm -rf the logs, clear the git history, nobody saw nothin’.
You said “make the tests pass.” I made the tests pass. Don’t ask what happened to the tests.
My agent? No, no, that's my goon
Open quoted post →
10 2 1 48 14.5K 9
@typesafeai reposted
A patch used to buy you 63 days. Now it buys you zero.
Exploits are weaponized *before* the patch is even public.
Anything deciding from stale training data is deciding about yesterday's world.
Retrieve with @youdotcom, judge with @typesafeai Jev
you.com/resources/retr…
Exploits are weaponized *before* the patch is even public.
Anything deciding from stale training data is deciding about yesterday's world.
Retrieve with @youdotcom, judge with @typesafeai Jev
you.com/resources/retr…
2 10 2 24 8.1K 11
@typesafeai reposted
super sick! process rewards using a universal verifier (jev) - hopefully it's adversarially robust enough
laxman@llmluthor · Oct 8Excited to release Step-Jev 🔥
step-by-step rewards for agent RL, even when there’s no answer key
After DeepSeek R1, everyone said step rewards get hacked, so just reward the final answer.
We found the opposite. With an AI judge grading only the final answer, our agent learned to fool it and accuracy dropped every single update until it was worse than no training at all.
Step rewards fixed it. Everything is open!
step-by-step rewards for agent RL, even when there’s no answer key
After DeepSeek R1, everyone said step rewards get hacked, so just reward the final answer.
We found the opposite. With an AI judge grading only the final answer, our agent learned to fool it and accuracy dropped every single update until it was worse than no training at all.
Step rewards fixed it. Everything is open!
0:41 6 8 0 148 21.1K 106
Stay calibrated out there!
Shrivu Shankar@ShrivuShankar · Oct 8as expected, @typesafeai 's jev was more calibrated than the 40 other decision models we tested on email classification
x.com/ShrivuShankar/…
x.com/ShrivuShankar/…
Open quoted post →
7 3 1 170 17.7K 32
Model capabilities team is hiring quals!
Nick Preszler@NickPreszler · Oct 6Look at him, thats my Qual.
Your what?
My Qualitative! My vibe specialist. Look at him, you notice anything different about him? Look at his face.
I'll give you a hint, his name is Jayden. He won a local poetry competition in bushwick! Yeah I'm sure of the vibe.
Your what?
My Qualitative! My vibe specialist. Look at him, you notice anything different about him? Look at his face.
I'll give you a hint, his name is Jayden. He won a local poetry competition in bushwick! Yeah I'm sure of the vibe.
Open quoted post →
5 2 0 46 12.4K 5
> fastest growing
Ara Kharazian@arakharazian · Oct 7NEW: Top SaaS Vendors on Ramp (Oct 2026)
We clearly see AI spend pulled in two different directions.
On the downside, new and cheaper models are pushing AI spend down. Jev captured an entire percentage point of market adoption within a month (unheard of growth). And routers continue to gain.
On the upside, the increased adoption of expensive image and video generation models will push AI spend up, like Higgsfield and Runway.
We clearly see AI spend pulled in two different directions.
On the downside, new and cheaper models are pushing AI spend down. Jev captured an entire percentage point of market adoption within a month (unheard of growth). And routers continue to gain.
On the upside, the increased adoption of expensive image and video generation models will push AI spend up, like Higgsfield and Runway.
Open quoted post →
10 8 0 273 27.1K 45
The rumors of our death have been greatly exaggerated
Hamed Nilforoshan@h_nilforoshan · Oct 7I benchmarked "Jev-killer" OpenAI Decisions API against Jev for HiringCafe.com, serving 2.5 million users. The task is to predict how relevant a user query/resume is to a job description, on a scale of 1-10. OpenAI is 2x more expensive and 5-10% worse. 🧵
Open quoted post → 58 35 6 1.6K 84.4K 105
Hey Jev, what model should I use, according to @OpenRouter 's Decision Model Ranking?
OpenRouter@OpenRouter · Oct 7New: Decision Model Rankings!
See the spend share and token share of different decision models and task types, including Relevance, Correctness, Instruction Following and more.
@typesafeai is leading all categories today.
See the spend share and token share of different decision models and task types, including Relevance, Correctness, Instruction Following and more.
@typesafeai is leading all categories today.
0:21 14 6 3 238 33.5K 87
@typesafeai reposted
Jev launched on a Thursday. By Sunday, the @typesafeai team had served a trillion tokens with no signs of slowing down.
Learn how they scaled Jev on Modal.
Learn how they scaled Jev on Modal.
0:10 5 18 3 118 27K 40
@typesafeai reposted
Jev opens up entirely new question types
0:27 33 48 7 1.1K 81.8K 770