@HammadTime
Account based inUnited States
About this account
- Account based in
- United States
- Connected via
- United States App Store
Account-level information from X, not a live location or the device used for a specific post.
normal considered harmful | cto @trychroma
Berkeley, CA
Joined September 2009
- Tweets2.6K
- Following2.8K
- Followers2.1K
- Likes25.8K
yeah....thats a no from me dawg
Surprisingly a robot can sometimes do highly-dynamic manipulation by just copying the demonstration 🤠. Without any learning, cracking a whip or lassoing a cleat can be done directly from human motion capture. And amazingly this process can be nearly automatically implemented by GPT-6 Astra
Check out the full blog post with fun visuals: krishnasuresh.org/blog/2026/…
Turn on sound for full effect 🔈
Quick side note: the whips shown in this post are called signal/stock whips. These whips are built to generate the loud cracking noise when the tip breaks the sound barrier and is a popular art form: youtube.com/watch?v=ietBtr7s….
I used to respect Yann but I think he’s finally lost me.
Yann LeCun: we are not going to get human-level intelligence or AGI by just scaling LLMs.
“There’s absolutely no way in hell.”
“The idea that we’re gonna have a country of geniuses in a data center, that’s complete BS.”
“It may feel like you have a PhD sitting next to you.”
“But it’s not a PhD.”
“It’s a system with gigantic memory and retrieval ability, not one that can invent solutions to new problems.”
😄
we’re excited to share that we’ve raised $20M (from Accel, Index, & Emergence) to build the 'ai-native Slack' you've all been craving
TLDR; ando is a team messaging platform built fresh, from the ground, up for a world where agents act as real collaborators alongside us.
we have teams using us across 15 countries, and spanning industries like software, financial services, real estate, & more
request access if any of this resonates 💌
Was great to hangout with @GregorVand on @software_daily
Hammad Bashir is the CTO of @trychroma. He joins @GregorVand to discuss AI retrieval, context rot, agentic search, small models, and the future of AI data infrastructure.
@HammadTime
softwareengineeringdaily.com…
the saddest part of AI writing my code is the depths to which I understand rust lifetimes to the point where I can construct lifetime cases that boggle the mind is now completely useless.
but alas, this is good, it should be useless.
hammad 🔍 retweeted
I see it as a point on the LLM pareto optimal curve in a regime that had a large revealed latent demand (no thinking, single token, low latency acceptable intelligence) that was under-invested into because of a race to higher intelligence.
important trend
LLM token output speed increases by 2-7x per year, with Fable-class models doubling every month (month!).
If trends continue, LLMs could meaningfully control robots in real time by end of the year, or by 2029 at the latest.
ALT Log-scale chart of LLM median output speed in tokens per second by release date, 2024 to 2026, with models binned by intelligence level. Each band's frontier climbs as a staircase: 2.3x per year at GPT-3.5 Turbo level, 7.8x per year at o1 level, 3.4x per year at GPT-5 level, and 2.1x per month at Fable 5 level. Data from Artificial Analysis.
people underestimate the degree to which large companies open sourcing software is about labor fungibility
so utterly tired of consciousness in AI conversations! especially when it comes to multi agent systems.
First, it’s not a binary! Go read @drmichaellevin
Second, It’s a terrible engineering tool. Better to ask what cognitive capabilities we are witnessing, .
plz stop pontificating and go be an engineer.
at-scale language model training is akin to semiconductor process development
1. small experiments lie, failures only appear at scale
2. you are fighting a yield problem
3. need world class observability into dynamics. how to do so is non obvious
the developed recipe compounds
maybe eval awareness is not to be squashed but harnessed
hammad 🔍 retweeted
- But zoom back even further, and a disagg WAL is simply an instance of Lamport's State Machine Replication, which inherently uses a shared, distributed log. (Our WAL-on-S3 design at Confluent heavily relied on this theory; as usual, it turns out Lamport got there first!)
morsel driven parallelism ftw
My favorite change we made to @pgrustdb is how we rearchitected parallel queries. Postgres decides a fixed degree of parallelism at plan time. For pgrust, we assign cpus to queries dynamically. For beefy machines with many cores, this results in a massive speedup
hammad 🔍 retweeted
We invented a new concurrency control protocol, named Fission, for agents swarms to update shared data for Foundation.
trychroma.com/engineering/tr…
and includes a section on "why not Git"
hammad 🔍 retweeted
Allow me to say that retrieval is the only ML problem in which we still cling to the assumption that an inert scoring function (a meager dot product!), which couples your “search-time compute” with the dimensionality of the representation, can be sensible.
Who would have guessed this forces you to have exponentially larger embeddings just to do basic things. Instead, if this makes it sound more modern, you need “inference scaling for retrieval”, also known as late interaction from 2019/2020:
Retrieval Needs Multivectors: An Exponential Separation
Microsoft formally proves that multi-vector embeddings can be exponentially more compact than single-vector ones for ranking documents.
📝 arxiv.org/abs/2608.21494
