Packed house @cognition's rooftop last night in SF - awesome to hear from @ScottWu46@breeves08 & Cognition team on state of AI code market, how to compensate exceptional people & Cognition's unique culture 🚀
We’re still early in closing the gap from AI adoption to capability (especially in non code verticals like customer support & legal) - because the reality of implementing & integrating AI in complex systems with proprietary data is hard!
“An application earns its place in the untrainable corner by doing unglamorous work: arranging a company's private reality so a model can act on it, handing the model the tools to act, working with the customer to change the reality of its workforce.” 👇
Neofirms will be defined by business models that charge for outcomes instead of hours. In order to win, they must measure the efficiency of labor + token spend, rather than competing to maximize token spend.
At Crosby, we use the Ramp AI spend dashboard to understand how we're spending tokens on customer outcomes, where efficiency = customer outcomes delivered / [token cost + legal cost].
The billable hour is optimized for the opposite, where more hours means more revenue. For us, every token and every hour of legal time comes out of our margin, so the pressure is always to deliver more with less. For the first time, the firm and the client actually want the same thing.
The firms that win won't be playing at the Token Casino. The firms that win will be the ones delivering the best outcomes at the lowest all in cost -- and measuring it closely enough to keep improving.
PSA: Just added a few thousand chips, including B200s and B300s to our Dedicated Model Inference (https://t.co/PiJhHYHVZh).
With Dedicated Model Inference, you can now on-click deploy our Blackwell optimized inference engine with auto-scale on frontier OSS models including Nemotron, Minimax, Kimi, DeepSeek, GLM and Qwen.
More than half of Cognition's first 50 employees were former founders.
@ScottWu46 sat down with @breeves08 and @lux_capital to talk about correctness over comfort, staying focused on our mission, and what it takes to build an incredible team.
one of the greatest founders Lux ever partnered with
REARDON
We first partnered in CTRL-Labs where he built coveted non-invasive brain-machine interface to CONTROL computers that Meta aggressively sought + acquired
NOW recruiting INSANELY smart talent🤯
unveiling: FLOURISH.
Another huge milestone for @Lux_Capital portfolio co @tryramp - explosive growth and one of the most interesting app companies shipping LLM features & infrastructure today 👇
Today, Ramp raised $750M at a $44B valuation.
Last time we grew this fast, we were 1/20th the size.
For 2000 years, business was built on two pillars. Today, a third: intelligence.
It’s your least governed cost. It’s also your single greatest opportunity.
Say hello to the AI operating system that learns your accounting firm's routine work, then autonomously runs it.
Reconciliations, journal entries, transaction coding, the whole close. Your team only touches what needs a human.
This is Stack. Build it once, run it forever.
our mantra is "any agent, any model"
devin started with built-in model routing, and then we expanded those learnings to devin managing devins
now, we're extending that to devin managing other agents too
Introducing Devin Desktop: the next generation of Windsurf
Manage fleets of local and cloud agents from one surface
Support for any ACP-compatible agent
With a full IDE for when you need to jump into the code
Another big launch from @cognition - Devin Desktop-
the best place for developers to manage the best agent for the job, bringing together Windsurf & Devin under one roof 💫
I think I was feeling the AGI much more 1 year ago than now. There’s just so much that’s broken and requires rebuilding in software for any decent real AI native application that will just not be vibe coded. Telemetry and analytics still separate (this is necessary for dark factory, and snowflake as sink is not going to cut it), filesystems with multiple writers + file locking that can be mounted anywhere, … In a private group chat, three founders said they built their own OTEL on top of @ClickHouseDB because they wanted agents to triage better while also save costs. Good lord we’re still early. Somehow, it seems like we will automate the scaling pilled form of AI research and math research far before customer support. It’s an insane world! Somehow magnificent and useless.
Mary Barra joined General Motors as an intern when she was 18 years old, and worked her way up from inspecting car fenders and hoods to becoming the CEO.
Please welcome @crosbylegal's INSANE summer intern class. I can't wait to see which one of them takes my job.
Reinforcement learning has exploded on Modal, and we've been cooking.
Here's a review of lessons learned helping teams train at scale, the patterns we kept seeing, and an open-source library to get started with RL on Modal quickly.
Startup builders: this is the NY Tech Week event you won't want to miss 👀
Join @UpstartsMediaCo as I interview 3 of the top startup founders in the Big Apple 🍎
▪️Modal CEO Erik Bernhardsson
▪️Merge CEO Shensi Ding
▪️Crosby CEO Ryan Daniels
Real talk about scaling and building an AI startup alongside real customers in the city that keeps it real. Thursday, June 4.
Apply here 👇
https://t.co/mOTjUPpOHQ
Minimax M3 is a remarkable model in many ways, but esp for inference efficiency. Minimax model designers and @togethercompute inference team will discuss it tomorrow live.