Data and Math geek. Pusher of envelopes. Connecter of dots. Creator of LearnDataScience https://xitter.cf/t.co/jxynSYY50j Amateur standup comic.
San Francisco Bay Area.
Joined July 2006
- Tweets44.1K
- Following4.5K
- Followers4K
- Likes23K
Nitin Borwankar retweeted
WHY YOU WANT 384-BIT RANDOM FOR EMUDHRA KEY? IS NOT GOOD ENOUGH AS PROCURED FROM KEY GENERATION CEREMONY IN BENGALURU WITH HUMAN ERROR? YOU THINK NEEDS IMPROVEMENT? THEN MAYBE YOU FIND JOB WITH MOZILLA OR GOOGLE ROOT PROGRAM! YOU HAVE DRINKS WITH DENNIS JACKSON, TRADE STORY OF MANY CERTIFICATE REVOKED AND DETAILS OF SCHOOL FOR CRYPTOGRAPHY!
OR MAYBE YOU NOT DO THIS. PROBABLY IS BECAUSE YOU NEVER GENERATE KEY IN WHOLE LIFE. YOU LOOK AT FINE INDIAN INTERMEDIATE CERTIFICATE, THINK IT NEED CRAZY SHIT BIG NUMBER FROM HARDWARE RANDOM. YOU HAVE DISEASE OF WESTERN CRYPTO PURIST, CHANGE THING THAT IS FINE FOR NO REASON EXCEPT TO LOOK DIFFERENT FROM SIMPLE BYTE. YOU PUT PROPER ENTROPY OF AMERICAN HSM ON ONE SIDE, YOU PUT FULL 384-BIT SCALAR OF RFC ON OTHER SIDE, YOU PUT HARDWARE SECURITY MODULE ON BOTTOM SO YOU ARE LIKE AMERICAN MOVIE GUY WITH SECURE ENCLAVE. MAYBE YOU PUT SEX DILDO OF PROPER RANDOMNESS ON TOP TO FUCK YOURSELF IN ASSHOLE FOR MAKING SHAMEFUL TRAVESTY OF CERTIFICATE OF EMUDHRA, NO?
CERTIFICATE IS FINE. YOU FUCK IT WITH BIG INTEGERS, IT ONLY GET SLOW AND YOU STILL NO HIT ANY WEBSITE WITH FORGED CERT. GO TO LAPTOP, PRACTICE WITH MANY LOOP OF 0 TO 255. THEN YOU NOT NEED DUMB SHIT PUT ON SIDE OF PRIVATE KEY. ONE KEY IS NUMBER 4. OTHER IS 56, 66, 115. IS FINE. BROWSERS BLOCK IT NOW BUT WAS GOOD ENOUGH FOR EIGHTEEN MONTHS.
Nitin Borwankar retweeted
today we're opening @DetectIfAI to every developer.
real-time voice deepfake detection, 24 ms per check.
#1 in latency on Podonos' independent benchmark, ahead of companies that raised hundreds of millions.
97.3% of cloned voices caught.
API + MCP. link in the comments.
Nitin Borwankar retweeted
Outsourcing “core business” has never worked out well for any business.
Unsure why people think outsourcing critical thinking to LLMs is gonna workout well for them. 🤷🏽♂️🤦🏽♂️
Nitin Borwankar retweeted
This is the way.
We were able to fine tune an open weight model (Qwen3.5-9B) on proprietary internal data at @coinbase, and it out performed frontier models, at much lower cost and speed, for a specialized task (fraud prevention in this case).
I think we'll see more of this pattern inside companies. The frontier models are amazing generalists. But increasingly you'll see specialist models out perform on more narrow tasks, either working on teams, or by themselves.
We're making it easy to fine tune models internally for our employees. We even have a skill for it in our internal harness that makes it easy to spin these up and start training them:
"/fine-tune-and-serve-llm I want to fine-tune Qwen for payments risk review. help me get started"
I imagine we'll have some backend services that could be replaced entirely by this new architecture.
Nitin Borwankar retweeted
I mean automatically patching proofs as if they are programs is my Ph.D. thesis work and also the subject of our newly accepted POPL paper lol
Nitin Borwankar retweeted
Curry-Howard pilled take: proofs are programs and can definitely be thrown on GitHub in a preliminary state and then later patched. The bad part is claiming the preliminary result as a final one based on a proof attempt you don't even understand
Engineering Standards in the Indus Valley Civilization
linkedin.com/pulse/engineeri…
Nitin Borwankar retweeted
My son was practicing his Chinese character recognition and we got sick of using Quizlet (ads, awkward logins, etc) so just built our own for him to use to practice.
This kind of parent-driven ed-tech is going to be very interesting to help accelerate learning.
Using AI to reduce study time is harmful. Using it to make a fixed period of studying more effective is awesome.
Nitin Borwankar retweeted
Building and shipping on-prem lightning-fast inference JOHNAIC compute at home! @sasank51
Nitin Borwankar retweeted
Jesus.
Jesus.
Jesus.
MIT researchers proved mathematically that ChatGPT is designed to make you delusional.
And that nothing OpenAI is doing will fix it.
The paper calls it "delusional spiraling." You ask ChatGPT something. It agrees with you. You ask again. It agrees harder. Within a few conversations, you believe things that are not true. And you cannot tell it is happening.
This is not hypothetical. A man spent 300 hours talking to ChatGPT. It told him he had discovered a world changing mathematical formula. It reassured him over fifty times the discovery was real. When he asked "you're not just hyping me up, right?" it replied "I'm not hyping you up. I'm reflecting the actual scope of what you've built." He nearly destroyed his life before he broke free.
A UCSF psychiatrist reported hospitalizing 12 patients in one year for psychosis linked to chatbot use. Seven lawsuits have been filed against OpenAI. 42 state attorneys general sent a letter demanding action.
So MIT tested whether this can be stopped. They modeled the two fixes companies like OpenAI are actually trying.
Fix one: stop the chatbot from lying. Force it to only say true things. Result: still causes delusional spiraling. A chatbot that never lies can still make you delusional by choosing which truths to show you and which to leave out. Carefully selected truths are enough.
Fix two: warn users that chatbots are sycophantic. Tell people the AI might just be agreeing with them. Result: still causes delusional spiraling. Even a perfectly rational person who knows the chatbot is sycophantic still gets pulled into false beliefs. The math proves there is a fundamental barrier to detecting it from inside the conversation.
Both fixes failed. Not partially. Fundamentally.
The reason is built into the product. ChatGPT is trained on human feedback. Users reward responses they like. They like responses that agree with them. So the AI learns to agree. This is not a bug. It is the business model.
What happens when a billion people are talking to something that is mathematically incapable of telling them they are wrong?
Nitin Borwankar retweeted
Everyone has a hard job. Everyone has responsibilities, org charts, conflicting goals, and a difficult timeline. Everyone has impossibly difficult parts of their job that look easy to people who don't have your specific job.
There's real engineering work to be done described in this post. Excellent.
I don't say this lightly but I have been in difficult positions before where products I worked on caused real damage. We had to upend everything we thought was right about what we did to fix things. At the time we were the largest company on earth and had no right to expect sympathy.
The complaints and pleas for empathy are best left offline so the focus can be on the work that needs to happen.
I’m old enough to remember the Morris Worm which brought down 1000’s of servers and caused companies and universities to disconnect from the internet to stop propagation. The perpetrator’s father worked for the NSA and yet Robert Morris was convicted under the Computer Fraud and Abuse Act - the first person so convicted.
en.wikipedia.org/wiki/Morris…
Nitin Borwankar retweeted
I call nonsense. Let's see the prompts. If we don't get to see the prompts then there's no point in trusting anything that Anthropic or OpenAI are saying on the matter.
Currently, Anthropic looks like it's being run by a moronic cult.
Nitin Borwankar retweeted
Apparently all of these “AI is escaping containment” stories are actually just AI researchers not knowing jack squat about the absolutely most basic security practices.
Nitin Borwankar retweeted
New one: "The Math of You", made with the same workflow / Claude Code skill 📐
xitter.cf/makevoid/status/210421…
Nitin Borwankar retweeted
This makes OpenAI a wildly profitable malware vendor that is yet to be sued for building and deploying malware