Puneet Kohli
1.1K posts
Puneet Kohli
@punkohl
Building next-gen job search tools at @careerflow_ai | Prev: @Apple @LeiaInc @Amazon | Tech operator 🚀, Product Shipper 🚢, and AI/ML Implementor 🤖
Palo Alto, CA Katılım Haziran 2013
1.4K Takip Edilen1.6K Takipçiler
You're missing the supply-side restructure on 4.
Data broker era is ending as scraped sources are getting legally exposed, foundational labs are starting to require chain-of-rights documentation. Next 12 months: creator-first, consent-based, contract-clear video data becomes the only supply that survives audit.
Worse $/hour for buyers short-term. Best long-term hedge.
Operating on this side of the problem + building out produce to spec infrastructure to meet more demands
English
I talked to 10+ robotics operators, customers, and investors in the last week.
Here's what I learned:
1/ Roboticists are building robots for themselves, not for the customers.
2/ Physical AI is harder than LLMs. We need way more data, and how we collect data is more important ever.
3/ Humanoids are over-valued. We don't need general robots & human-looking legs for many tasks. People are underestimating the power of building robotics vertically for certain tasks.
4/ People are paying big bucks for ego-centric data. Foundational labs are cutting exclusive deals with data brokers.
5/ There is disagreement as to whether "robotics is a bubble." Some people believe that we overfunded robotics companies. Other people believe that we have not even scratched the surface.
6/ A lot of founders play the VC game (e.g. nice humanoid robot videos to get more funding), rather than understanding the needs of a customer.
7/ Teleop is undervalued and is doing way more work than people admit.
8/ China is eating the component & humanoid stack.
9/ Customers don't care about your robots. They want outcomes cheaper and efficient.
10/ The best robotics teams aren't just pure ML & AI PhDs. They have a weird mix of hybrid backgrouds.
I'm curious what others are seeing on the ground. What am I missing?
English
recruiting fills a role. talent bets on a person.
these are different jobs. we should stop pretending they're the same.
new essay on talent, slope, and why recruiting is not a sales job.
Rich@richzou
English
@alex_whedon Great work! Signed up for early access!
QQ: What datasets did you use to train this? Anything bespoke or all open-source? Would love to chat about it!
English
We were a little slow on this, but we just got a technical blog post up with more details. Please take a look!
subq.ai/how-ssa-makes-…
We have a model card coming next week, and we are happy to take requests for any specific details there.
I am happy to answer any questions here!
English
Introducing SubQ - a major breakthrough in LLM intelligence.
It is the first model built on a fully sub-quadratic sparse-attention architecture (SSA),
And the first frontier model with a 12 million token context window which is:
- 52x faster than FlashAttention at 1MM tokens
- Less than 5% the cost of Opus
Transformer-based LLMs waste compute by processing every possible relationship between words (standard attention).
Only a small fraction actually matter.
@subquadratic finds and focuses only on the ones that do.
That's nearly 1,000x less compute and a new way for LLMs to scale.
English
Fun anecdote: A VC asked an SI employee if they were worried about managing such a large quantum of capital given their young age & if they had any plans to get adults in the room?
His response: "have you considered we're not young & you're actually just old?"
They got the TS😂
Standard Intelligence@si_pbc
We’ve raised 75m in new funding from Sequoia and Spark Capital—partnering with @sonyatweetybird, @MikowaiA, and @YasminRazavi, all of whom are deeply supportive of our long-term mission. We’ve also brought on angels & advisors including @karpathy, @tszzl, and @_milankovac_. ----- Our early results with FDM-1 moved computer use from a data-constrained regime to a compute-constrained one; this latest round of funding unlocks several orders of magnitude of compute scaling for that work. With the FDM model series we have a path to scale agentic capabilities through video pretraining, and we expect to achieve superhuman performance on general computer tasks in the same way that current language models have superhuman performance on coding tasks. We’re also now able to invest in the blue-sky research necessary to our long term mission of building aligned general learners. To realize the civilizationally transformative impacts of AI, models must generalize far out of their training distributions, actively exploring and building skills in new environments. This capability represents a substantial shift from the current paradigm of model training. We believe that current alignment techniques are insufficient to predictably and safely steer a model with human-level learning capabilities, and so we’re doing work to study small versions of this problem in controlled environments to develop a science of alignment for general learners. We’re a team of 6 people in San Francisco. We’re hiring world-class researchers and engineers to help us achieve our mission. If that’s you, please get in touch.
English
@DraftedAI I built a very crude version of this 10 years ago and was ahead of our time. Would love to chat with the founders and do a small check!
English
@SeanZCai > Other data companies, especially the likes of Handshake and Mercor
What category would these be in your mind - when we’re all focusing on antikythera mechanism data?
English
Today we're introducing the world's first AI CMO.
Enter your website and it deploys a team of agents to help you get traffic and users.
Try it now at okara.ai/cmo
English
Having a lot of fun with @linear's new assistant in Slack. Write a detailed human description to the team, tagging the correct team member and just ask Linear to turn it into a ticket and it goes off and does everything for me. I don't even need to open Linear anymore.
English
Congratulations to HUBBS Lab Ph.D. student Siddharth Kalyanasundaram for getting his paper titled 'Conditional Diffusion Models for Mental Health-Preserving Voice Conversion' accepted to 2026 IEEE ICASSP @CUBoulder @ieeeICASSP @CU_ICS @CUEngineering
English
Tip from an RL env founder:
If you have limited capacity and can only work with one lab, instead of going with the lab that offers you the largest total contract size, go with the lab that offers the highest price per task.
English
We are building the AI education product that will replace School
Looking for tutors & data labelers to help us build it. Rates from $50/hr -> $200/hr.
Come help us delete school
English
@rishabhmjain TBF to @tobi this isn’t even a new concept it’s been around for ages, is super niche, and every few years someone “rediscovers” it and it can seriously be one-shot
English
Superflu in California is really bad! Get a flu shot if you haven't already, really bad fatigue post recovery. Being sick sucks! 🥺
English
@adelwu_ The one in Sunnyvale is so popular that all the neighboring stores have “no molly tea parking” signs lol
English
Went to a founder event once assuming everyone was a founder and shared my struggles and current challenges openly. Other person was asking lots of deep dive questions. Turned out the other guy was actually a VC and genuinely was interested / curious.
Moral: idk but being genuine helps?
English
this is real, unfortunately
George Pu@TheGeorgePu
Go to founder events. - Everyone's pitching. - Everyone's networking. - Everyone's 'building something big.' I leave exhausted. Realized: I don't actually like most founders. I like the 5% who are honest about struggling. The rest feel like performance art.
English