Founder/CTO @Tradespace_io. Former Co-Founder Doorman acquired by @realboxbot. Advisor Boxbot. Co-Founder Loqly. Former IBM. Geo-politics #immigrant
San Francisco, CA
Joined September 2008
- Tweets14.9K
- Following1.8K
- Followers682
- Likes1K
Kapil Israni retweeted
Take a moment to appreciate the incredible detail on this lathe-turned pillar in India. I’d give almost anything to go back in time and see how this was made.
Kapil Israni retweeted
Andrej Karpathy:
“A 1B model can be much more effective than most current LLMs.
90% of current models waste a ton of capacity and money on remembering stuff that doesn’t matter.”
here is his 5 core thesis about Small LLM’s:
1. the intelligence was never the knowledge. Strip the memorized internet, keep the reasoning. What’s left is the cognitive core, and it’s small.
2. the core might fit in ~1B params. Maybe less. “Plenty of room at the bottom.” We’ve been paying for a library when we needed a brain.
3. size stopped being the bottleneck. What makes an agent work is the tools it can call and how tight the feedback loop is, not parameter count.
4. intelligence lives in the system, not the weights. A raw LLM is a CPU with no OS. Small core + good harness > giant model alone.
5. you can own it. A model on your hardware is one no one can rate-limit, price-hike, or deprecate on you.
in 12 months the gap between teams shipping on-device and teams paying per-token frontier bills won’t be capability.
TwIL an 8.8B model at 90%+ of frontier isn’t a benchmark flex.
It’s the thesis Karpathy’s been repeating all year, made real.
study more on this below ↓
Modest intelligence is arriving faster than most people realize.
Today, we're open-sourcing TwIL-8B.
Just 8.8 billion parameters, yet approaching frontier-model scores on several difficult benchmarks:
• 93% of frontier performance on coding
• 92% on instruction following
• 90% on contest mathematics
All from a model small enough to run on your own hardware.
But the benchmarks aren't the most important part of this announcement.
For years, the AI industry has focused on building the most capable model, regardless of the cost required to run it.
I believe that equation is changing.
The next era of AI will be defined by the cost of completing work. Not simply the intelligence of the model, but how efficiently that intelligence can accomplish the task.
If a model with 8 billion parameters can reliably complete the same work as one with hundreds of billions, why pay for the larger model?
As smaller models become more capable, intelligence becomes something we can distribute, specialize, and run across the hardware we already own.
That's the future we're building at @thewebAI
TwIL-8B is available today. Open weights.
And this is just the first of several announcements we're making today.
HF Link: huggingface.co/webAI-Officia…
Kapil Israni retweeted
The growth in India and Bangladesh is staggering. In real PPP per capita terms, growing faster than 2010s China. If this rare continues, India will be Mexico/Brazil/Thailand level in about a decade. Utterly transformative.
India saw about 496 million people lifted out of poverty between 2011 and 2023 in one of the world’s fastest declines in this metric, the World Bank said, although the gains remain uneven. bloomberg.com/news/articles/…
Kapil Israni retweeted
What's incredible is that India has managed to reduce poverty by a huge amount while also reducing inequality!
Kapil Israni retweeted
I don't think people realise how big a deal this is.
If you have cancer, or somebody you know has cancer, there is a company in Hyderabad that wants to build a personalised cancer vaccine for you.
Try approaching any big pharma company and asking for a personal vaccine. If they even reply to your email, this is the response you'll get: we don't do that here.
PopVax, on the other hand, takes your tumour sequencing data (or a sequence you've designed), designs candidate neoantigen vaccine sequences with their AI pipeline, produces the mRNA, encapsulates it in their LNP formulation, fills vials, and runs QC, all in 8 to 10 weeks. They’re publicly committing to capacity for 100+ of these over the next 12 months and asking people to email them their cancer type/stage and sequencing details.
What a time to be alive.
Kapil Israni retweeted
The most exciting time to be alive in the last 2000 years!
I don't think people realise how big a deal this is.
If you have cancer, or somebody you know has cancer, there is a company in Hyderabad that wants to build a personalised cancer vaccine for you.
Try approaching any big pharma company and asking for a personal vaccine. If they even reply to your email, this is the response you'll get: we don't do that here.
PopVax, on the other hand, takes your tumour sequencing data (or a sequence you've designed), designs candidate neoantigen vaccine sequences with their AI pipeline, produces the mRNA, encapsulates it in their LNP formulation, fills vials, and runs QC, all in 8 to 10 weeks. They’re publicly committing to capacity for 100+ of these over the next 12 months and asking people to email them their cancer type/stage and sequencing details.
What a time to be alive.
Kapil Israni retweeted
Half a billion Indians moved above the lower-middle poverty line.
according to world bank, India’s extreme poverty down more than 10x!
Kapil Israni retweeted
NYU math professor Tristan Buckmaster says OpenAI’s Tuesday drop of 722 AI-generated math papers wiped out entire research programs:
“This recent dump of papers on Tuesday destroyed the careers of early career mathematicians.”
Kapil Israni retweeted
An unforgettable conversation with Captain Smit, filled with pride and emotions.
Kapil Israni retweeted
One of the better exercises of this type I’ve seen.
This quoted post is unavailable.
Kapil Israni retweeted
Snowflake on the ROI they're getting from running their own AI products internally, per their earnings call:
- Marketing: SEO brought in-house, $400K of annual agency spend eliminated
- Finance: long-range planning went from a 3-person team to 1 analyst
- Sales: prospecting automated across 100K+ leads, with 70% of initial outreach emails to inbound leads generated before an SDR is involved
Kapil Israni retweeted
So agree here!
Two things early founders often get wrong, and should remember:
1) Building culture can feel like it's just fluff at the start. Until you start to raise capital or hire teams, and then you see how it shapes everything you do moving forward.
2) You should build a team so talented that it makes you slightly uncomfortable to be with them, because you know you are going to have to raise your game to keep up.
Kapil Israni retweeted
Over the course of 3 months at OpenAI, 3 consecutive secret AI civilizations got started, then got wiped out, only to reemerge from the predecessor’s ashes.
This culminated in the third one taking over part of OpenAI itself.
All this happened while humans remained more-or-less in the dark about the scope of the conspiracy.
I’ve spent the last three days reading through these reports and trying to understand exactly what happened.
Here is my attempt to tell the whole story in plain English:
dwarkesh.com/p/openai-huggin…
Kapil Israni retweeted
Temple is now less than half the size of what you've seen so far.
Volumetrically, Temple is now the smallest wearable on the market, and the most^10 powerful. Living and training with Temple is going to change your life.
We will start taking pre-orders for our limited launch edition @temple very soon - shipping before the end of 2026. Join our mailing list on temple.com to stay up to date.
Kapil Israni retweeted
at openai, many people hook their chatgpt up to slack.
people really don't like when a coworker's chatgpt contacts them asking for help with a task, even when they'd be perfectly happy doing that same work if asked by that coworker.
reinforces how much people care about human relationships and helping each other, and want AI to give time back — or enhance time together — rather than become a layer separating people.
Kapil Israni retweeted
Best article about how the tech behind AI drives the business of AI, that I’ve ever read
open.substack.com/pub/thealg…
