frontier of your fav frontier lab/ founder @egma_ai / fellow @spc / prev founder @DoWhileAI (sequoia surge 09) / @iitroorkee
Joined March 2020
- Tweets4.3K
- Following2.1K
- Followers2.3K
- Likes13.9K
Pinned Tweet
Launching Egma - OSS simulation testing for voice agents
We've deployed voice agents in regulated industries ourselves & spent hours manually testing on every push.
Existing vendors charge an insane 5-20X premium on inference. Egma lets you bring your own API keys.
Github👇
day 1/n of instinct created explainer videos - grounded in designing data intensive applications book. i've always wanted to read it but never really got around to it. lets see how this expt pans out.
it'll send me a video each day - scoped one per sub-chapter. today - what is a data system?
i liked the zero shot quality!
I described in natural language to the new local iPhone AI to create a shortcut opening hackertube in safari & created a logo via chatgpt. wall time from seeing on X to my homescreen was <3 mins.
we live in wild times
Weekend hack: HackerTube. I was annoyed with YouTube's recommendations so I had Claude filter down 43M comments over 20 years of HN to 6,454 videos that keep being recommended. *Significantly* better viewing experience, ~2 hours with Opus 5.5:
hackertube.org/
- Excellent recs without any logins (state is kept in local browser session)
- Thumbnail is a representative frame from middle of the video instead of the super duper A/B test hyper-optimized user uploads
- Underlying data from the comment analysis as one zip file, point your Claude at it and ask for your own version
- HackerTube Longs ✊
A bit more on the process:
I use Claude Code (desktop app for Mac), Opus 5.5 on Medium effort as my default. I only talk to Claude and have it do everything, the only "manual work" I helped with in tandem was in the browser (reminiscent of MenuGen): 1) I had to create some kind of API key from Google Cloud UI to query YouTube for video metadata, 2) I booted up a VPS box for serving and 3) I bought the domain name. Once the box was up I gave Claude the IP and told it update, secure, harden, deploy and optimize the code, and it did the whole thing.
Many awesome projects that would have normally taken weeks/months even can now be few-shotted for breakfast. What else are we not thinking of.
increasingly doing this. i dont want your agent. i want you to be a data store. that's it.
at the limit - *most* saas should charge me for compute and storage
nischal jain retweeted
This is a LP letter that we recently wrote with our lessons on spotting talent from a decade of investing at the frontier.
docsend.com/view/p7wm6mezzbq…
The most common thing you hear about early stage investing is that "It is 90% about the founder".
We agree but have always been unsatisfied by the lack of falsifiability, specificity, and repeatability of that statement.
What are the dimensions of excellence? How do they translate into companies?
It was a pleasure to collaborate with @gopalkraman on this.
wow
I've always believed that intelligence should come from simple, general building blocks that learn their own abstractions directly from raw data; in other words, the bitter lesson. I think we're finally seeing a real step towards AI that can learn from all data.
Language modeling on raw bytes is a microcosm of this ideal, and to me it's a central problem in AI: if we can't build LLMs that learn end-to-end from bytes, how will we handle the rest of the world's data that we understand even less?
Last year we introduced hierarchical networks (H-Nets) as a first step. Our new version finds up to 3-5x efficiency gains in controlled scaling law experiments, and at larger scale matches an open-source model trained on 11T tokens with ~10x less data and compute. The posts below walk through what we found and why we think it works.
nischal jain retweeted
Today we're launching @tryAlanAI so developers can stop babysitting their coding agents.
We kept running into the same problem ourselves.
Even as agents got better at writing code, we were becoming glorified button-pushers for our own AI: starting every run, feeding them context, unblocking failures, and checking their work.
We built Alan to change that.
nischal jain retweeted
When maximally ambitious builders meet a moving frontier, just 18 months is all it takes.
To change everything.
Coming soon.
SPC India's Annual Portfolio Showcase.
Invite only. Watch this space.
is there an mcp/skill that securely exposes what passwords have been compromised or repeated to personal agents - which can then go and reset password across bunch of websites.
consumers need to up their password game and really make sure to have unique password for each site
i remember reading about this attack class few years ago and i was in awe of the folks that first figured it out.
but i was like meh this is just interesting. would require months or years of dedicated team work and maybe espionage to exploit it.
but damn. ai has changed it.
Agents can play "Ouija" in the cloud.
This class of timing attacks on shared-host VMs has been known since at least 2015¹. With no full solution.
The world’s cyber infrastructure runs on “yes, that attack is theoretically possible, but who would have the ingenuity & gall to go that far!” Enter, agents.
I'd describe this particular OpenAI find as a 'noisy neighbor' attack. If two agents are on the same machine, one can make enough noise (through the CPU caches!) that the other one can detect it by timing its own computation and cache accesses.
The only "complete" solution would be for each agent to run on its own dedicated hardware, which obviously has significant economic consequences.
For Vercel Sandbox, we've deployed every practical mitigation possible to this class of attack, complementing it with always-on monitoring and red-teaming (similar to where OpenAI arrived).
¹ ieeexplore.ieee.org/document… (and even further back)
nischal jain retweeted
new fav personal assistant use case unlocked!
i was annoyed that the roads near my house in HSR were dug up for MONTHS making it hard to even walk and it felt like there's nothing i could do about it!
early sept i asked faff to find a way to flag this to the authorities...
hackers did
“Who touched prod?” Now with filters
Changelog #0312
• Activity feed filters
• Guided IaC migration
• Cron jobs in the CLI
• Fixes and improvements
railway.com/changelog/2026-1…
why do social media sites not have a dislike/ downvote button to express disagreement @nikitabier
we're reinventing wine and dine from first principles
F your AI b2b SaaS.
We just launched Palantir for booking meetings.
The last 14 days, companies like @MonacoGTM , @Nowadaysai, @Rippling , @salesforce , @marketercom, @posthog , and @retellai spent $100k+ with us on literal cakes.
From today, Daymaker finds your prospects, reads their entire digital footprint, buys the one gift that makes their day, and gets you the meeting.
Comment your name and I'll reply with everything Daymaker finds on you and the gift it would send you. If you dare. 🎁
i think this is the final nail in coffin. very likely railway could be the first startup to go out of business because of ai driven cybersec incidents eroding customer trust on a infra provider. this is very very bad
blog.railway.com/p/security-…
