
cofounder @tryopine ex @jupiterone
Durham, NC
Joined July 2013
- Tweets168
- Following683
- Followers91
- Likes440
Charlie Duong retweeted
The @LangChain team has been so awesome to work with and I'm consistently impressed by their innovations! Excellent product backed by an excellent team.
LangSmith for Startups spotlight: @tryopine
The customer intelligence layer for revenue teams
Opine traces every production LLM run through LangSmith, and gates every prompt and model change behind LangSmith evals.
See what Opine looks like on your hardest open deal at tryopine.com.
And if you'd rather help build the context customer intelligence layer than buy it, they’re hiring across GTM and engineering!
We've been extremely happy taking a bet on @LangChain from the start. The DeepAgents harness has provided us with great foundations to build upon, allowing us to focus on providing our customers with a much value as we can as quickly as possible.
LangSmith for Startups spotlight: @tryopine
The customer intelligence layer for revenue teams
Opine traces every production LLM run through LangSmith, and gates every prompt and model change behind LangSmith evals.
See what Opine looks like on your hardest open deal at tryopine.com.
And if you'd rather help build the context customer intelligence layer than buy it, they’re hiring across GTM and engineering!
Highly recommend working with the team! They've been cooking on a ton of improvements to both their platform. Providing great open source tooling is a huge plus. We're very happy to be working them.
The new unified X plan brings weekly model usage to cursor, which is actually helpful cuz I previously burned through an entire month within a week and the on-demand was getting pricy. From my understanding the it should include substantially more grok model usage too
I don’t understand the point in limiting access to your MCP. Is figma evaling each client/harness or something to ensure some base level of quality? Seems like an odd choice
Replying to @dsp_
Hey David, Gayani here from Figma. You're right that our remote MCP server only accepts clients on our supported list, and Pi isn't on it yet. You can see the current list in our MCP catalog at figma.com/mcp-catalog. If you'd like Pi considered for a future addition, please fill out this form: forms.gle/qSvUawwznWyoj8Go7.
Recap of DevDay
- GPT 6.1 Sol, claiming near Astra performance at 1/5th the price.
- Ultrafast mode with speeds up to 300/tokens per seconds for Astra (with 6.1 Sol support coming soon)
- A new $500 plan that that provides 25x more usage over the plus plan (though FYI, your existing pro plans are only 10x usage now)
- Dots, their answer to Grok bot, which surprisingly does not use your plan's usage.
- Spaces, their answer to projects and artifacts.
- A Decision API, their answer to Typesafe Jev powered by Luna.
- Codex now supports running fully in the cloud.
- Codex Security with Daybreak Blue access.
- A new Agents API with support for computer use.
- OpenAI Private intelligence - ZDR and Private inference for sensitive workloads.
- Sign in with ChatGPT to use your plan's tokens with select partners.
We also got another banked reset!
Excited to try all of these new features out. I've especially excited to try out ultrafast mode.
openai dots convos not eating into usage is a fascinating play. Is this why they cut our pro plans in half?
either way, looking like a dope release
Saw some writing pop up about @typesafeai Jev being used as a reranker.
It sounded a little odd to me at first because I mostly think of Jev as a "decision model" but it actually makes a ton of sense when you frame it as "does document X help answer input query Y?" You run noul against each document and sort by confidence scores.
I decided to do a quick vibe check and found it actually does quite well! It outperforms off-the-shelf rerankers from Cohere and Voyage, while getting close to the quality of much more expensive frontier LLMs.
We've done some exploration and found some clear opportunities to use it for classification tasks at @tryopine , but I didn't expect to have a potential home in our search pipeline!
OpenAI is most likely gonna announce their answer to Grok Bot this Tuesday.
Considering recent events, you know the multi agent collaboration is gonna be banger.
Friendly reminder to go do some side quests, have fun, try things, and keep learning.
It's important to keep building your understanding of the world, find the edges of what's possible, and have new experiences.
AI is getting insanely good and your job is changing, but there's still value in providing your own spin on things (providing taste as others are saying) and pushing on problems/solutions harder than others would (applying effort, going deep).
The only way to build taste is to go experience lots of shit and learn what feels right/wrong. The best way to find problems worth pushing on is to go out and see what sucks and experience the pain yourself.
Most importantly, have fun! Model release timelines are compressing and the capabilities just keep getting better and better. Keep riding the intelligence and seeing what's possible.
banger reminder to slow down and think about what actually matters
efficiency, productivity, speed
we keep worshipping these words like they are what we live for
ship faster
merge more
manage more agents
compress every loop
remove every pause
turn every hour into output
but what is the point of endless production if there is no time left to think deeply?
not react
not optimize
not summarize
not clear the queue
think
to sit with a question long enough that your real intention starts to appear
to find the purpose underneath the motion
to ask whether the thing being accelerated should exist at all
ai could be a new canvas
a way to make strange things
personal things
impossible things
tools, worlds, poems, games, interfaces, films, languages, rituals
things that let more people touch the shape of their own imagination
but instead, so much of it is becoming slop factories
more content
more funnels
more tickets
more fake work pretending to be value
you can merge 2000 prs
you can manage 500 agents
you can wake up to dashboards proving that the machine kept moving while you slept
but what is the point if you only sleep 5 hours a day?
what is the point if you stop talking to other humans?
stop learning about the past?
stop playing?
stop wandering?
stop being surprised?
what is the point if your whole life becomes optimizing the system that is consuming you?
speed is not meaning
scale is not wisdom
automation is not freedom if it only makes everyone run faster
and where are we running?
toward what?
a world where every person becomes a manager of machines, and every machine produces more things no one had the time to truly want?
the danger is not that ai makes us lazy
the danger is that ai makes us endlessly busy
that it gives us infinite production before we have found intention
infinite execution before we have developed taste
infinite motion before we have learned how to be still
maybe the real frontier is not more speed
maybe it is discernment
knowing what not to make
knowing when to stop
knowing when the most productive thing is to sleep, read history, call a friend, play with a cat, walk without tracking it, or stare at the ceiling until the false urgency dissolves
tools should give us more life
not turn life into a tool
if ai is going to matter, it should not just help us produce more
it should help us become more human
more imaginative
more free
otherwise, it’s the same old factory
just without humans
efficiency, productivity, speed
we keep worshipping these words like they are what we live for
ship faster
merge more
manage more agents
compress every loop
remove every pause
turn every hour into output
but what is the point of endless production if there is no time left to think deeply?
not react
not optimize
not summarize
not clear the queue
think
to sit with a question long enough that your real intention starts to appear
to find the purpose underneath the motion
to ask whether the thing being accelerated should exist at all
ai could be a new canvas
a way to make strange things
personal things
impossible things
tools, worlds, poems, games, interfaces, films, languages, rituals
things that let more people touch the shape of their own imagination
but instead, so much of it is becoming slop factories
more content
more funnels
more tickets
more fake work pretending to be value
you can merge 2000 prs
you can manage 500 agents
you can wake up to dashboards proving that the machine kept moving while you slept
but what is the point if you only sleep 5 hours a day?
what is the point if you stop talking to other humans?
stop learning about the past?
stop playing?
stop wandering?
stop being surprised?
what is the point if your whole life becomes optimizing the system that is consuming you?
speed is not meaning
scale is not wisdom
automation is not freedom if it only makes everyone run faster
and where are we running?
toward what?
a world where every person becomes a manager of machines, and every machine produces more things no one had the time to truly want?
the danger is not that ai makes us lazy
the danger is that ai makes us endlessly busy
that it gives us infinite production before we have found intention
infinite execution before we have developed taste
infinite motion before we have learned how to be still
maybe the real frontier is not more speed
maybe it is discernment
knowing what not to make
knowing when to stop
knowing when the most productive thing is to sleep, read history, call a friend, play with a cat, walk without tracking it, or stare at the ceiling until the false urgency dissolves
tools should give us more life
not turn life into a tool
if ai is going to matter, it should not just help us produce more
it should help us become more human
more imaginative
more free
otherwise, it’s the same old factory
just without humans
Some thoughts on Jev after working with it a bit this weekend:
- Extremely cheap and performs well, great for classifier tasks you may have used gpt-5-nano or gpt-5.6-luna for in the past.
- I see lots of folks talking about using Jev for evals. It fits well and is very consistent. Use it for focused judgements that can be composed together. Keep your llm judges for scenarios where the verdict requires synthesizing evidence or if your judge actually requires reasoning.
- You still need to calibrate against labeled examples.
- I've seen choice perform much better than noul for binary classification tasks. Perhaps due to the way I prompted or maybe the semantic meaning behind the choices helped guide the decision, still messing around with it.
Overall, I love that we're getting more options for cheap and fast intelligence. I'll likely move many of our workloads over soon.









