Nitin Borwankar

Data and Math geek. Pusher of envelopes. Connecter of dots. Creator of LearnDataScience Amateur standup comic.

[Nitin Borwankar @nitin]

“Are not required” in this model of scientific endeavor. Does that mean only this model will produce science in future? Remains to be seen.

Say there is a regional problem in a third world country and it requires new science and engineering. But the private model is not available.
Does locally relevant science and engineering not “get done”?

This is a transitional phase where old problems are being rapidly solved but what comes next is unclear.
Having said that there is a massive privatization in progress no doubt.

[Nitin Borwankar @nitin]

@jmilinovich So genuine question - what exactly did Project Glasswing snd tge Mythos models achieve if critical infra is still so wildly vulnerable?

[Nitin Borwankar @nitin]

@officialKrishD Not to mention the massive audience it would draw.

[Nitin Borwankar @nitin]

If you look in any other industry it goes from advanced craftsmanship, where we are in SW, to industrial production.

But software is held back because we “craftsmen” and “artistes” do not want to let go.

Until today we were gatekeepers of progress because only *we* knew how to develop software and if we wanted to keep it at the level of craftsmanship that’s where it stayed.

Not anymore.

Someone or some entities with a clear vision and taking lessons from other industries can drive the transition to industrial production.

And we the craftsmen will be bypassed and left wondering what happened.

[Trash Panda 🦝 @trashpandaemoji]

@GergelyOrosz https://t.co/g1fgbrAwqi

[Nitin Borwankar @nitin]

@JhaSanjay There’s another one “It’s hard to get anything done in a Democracy”
Too many undertones in this one.

[Nitin Borwankar @nitin]

@TheKing18610116 @wideclops @GergelyOrosz A number between 1 and 256 does not have 384 significant bits.
At most 8 significant bits vs 384 needed.

[Tarini Sai Padmanabhuni @tarini_axory]

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.

[Neeraj Mathur @neeraj]

@TrungTPhan @hnshah 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 @nitin]

@andymstone Genuine question - what other known non-bot accounts has this happened to in recent times?

[mel 🫀☁️ @melodaysong]

getting absolutely wrecked by this today

[Nitin Borwankar @nitin]

@nickcammarata Does your work characterize in any way the topology of the underlying space, as high dimensional as it may be?

[Nitin Borwankar @nitin]

Next step of course is local models at a company distill the vertical knowledge out of the mega model and run a private SaaS app tuned for that company - run multiple such models but all based off a common data model. Hence fragmentation is minimal.

Is the mega model an evolutionary endpoint or a transition to, a waypoint for, a universe of custom local models?

[Nitin Borwankar @nitin]

@felixrieseberg Here’s a project where I had some initial thoughts and the rest was done by Claude via questions and research.

https://t.co/T46kT93L8a

[Steven Sinofsky @stevesi]

This is the way.

[Talia Ringer 🕊🪬 @TaliaRinger]

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

[Talia Ringer 🕊🪬 @TaliaRinger]

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

[Nitin Borwankar @nitin]

Engineering Standards in the Indus Valley Civilization

https://t.co/v45r1iLbKl

[David E. Weekly @dweekly]

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.

[Anshul Bhide @anshulbhide]

Building and shipping on-prem lightning-fast inference JOHNAIC compute at home! @sasank51