Vagmi Mudumbai

Make stuff with code

[PsudoMike 🇨🇦 @PsudoMike]

1/ Urai AI is out of Fredericton, New Brunswick. Two founders, two employees, building AI agents that pick up the phone and actually book the appointment, not just chat about it. They beat out more than 80 applicants for a New Brunswick Innovation Foundation seed check.

[JER @lifeofjer]

https://t.co/ofucbVgkLV

[Andrej Karpathy @karpathy]

A number of people are talking about implications of AI to schools. I spoke about some of my thoughts to a school board earlier, some highlights:

1. You will never be able to detect the use of AI in homework. Full stop. All "detectors" of AI imo don't really work, can be defeated in various ways, and are in principle doomed to fail. You have to assume that any work done outside classroom has used AI.
2. Therefore, the majority of grading has to shift to in-class work (instead of at-home assignments), in settings where teachers can physically monitor students. The students remain motivated to learn how to solve problems without AI because they know they will be evaluated without it in class later.
3. We want students to be able to use AI, it is here to stay and it is extremely powerful, but we also don't want students to be naked in the world without it. Using the calculator as an example of a historically disruptive technology, school teaches you how to do all the basic math & arithmetic so that you can in principle do it by hand, even if calculators are pervasive and greatly speed up work in practical settings. In addition, you understand what it's doing for you, so should it give you a wrong answer (e.g. you mistyped "prompt"), you should be able to notice it, gut check it, verify it in some other way, etc. The verification ability is especially important in the case of AI, which is presently a lot more fallible in a great variety of ways compared to calculators.
4. A lot of the evaluation settings remain at teacher's discretion and involve a creative design space of no tools, cheatsheets, open book, provided AI responses, direct internet/AI access, etc.

TLDR the goal is that the students are proficient in the use of AI, but can also exist without it, and imo the only way to get there is to flip classes around and move the majority of testing to in class settings.

[DHH @dhh]

If money couldn't buy me the freedom to work on Omarchy for the sheer love of computers, what good would it be?

[Vagmi Mudumbai @vagmi]

I am afraid that AI API providers are going to go this way.

[Rob Hallam @robj3d3]

Hey @nikitabier when will you stop gatekeeping the X API?

$200->$5,000/month is absurd.

We're not just losing the API. We're losing the next X add-on, the next analytics tool, the next thing that makes X better.

That's the real cost of gatekeeping your API. Is it worth it?

[Vagmi Mudumbai @vagmi]

I thought this would be about Rope and post training context extension. But this is much better. An awesome post.

[Avi Chawla @_avichawla]

The growth of LLM context length with time:

- GPT-3.5-turbo → 4k tokens
- OpenAI GPT4 → 8k tokens
- Claude 2 → 100k tokens
- Llama 3 → 128k tokens
- Gemini → 1M tokens

Let's understand how they extend the context length of LLMs:

[Jack Morris @jxmnop]

curious about the training data of OpenAI's new gpt-oss models? i was too.

so i generated 10M examples from gpt-oss-20b, ran some analysis, and the results were... pretty bizarre

time for a deep dive 🧵

[Madhu Guru @realmadhuguru]

At @Google, we are moving from a writing‑first culture to a building‑first one.

Writing was a proxy for clear thinking, optimized for scarce eng resources and long dev cycles - you had to get it right before you built.

Now, when time to vibe-code prototype ≈ time to write PRD, PMs can SHOW not tell. Role profiles are blurring, creativity and building are happening in parallel.

[Vagmi Mudumbai @vagmi]

Who the hell makes these decisions at Microsoft? Did they turn the HR department into a bunch of AI agents?🤯🤦🏽‍♂️

[Ron Buckton @rbuckton]

After 18 years at Microsoft, with roughly a decade of that time working on TypeScript, I have unfortunately been let go in the latest round of layoffs. I need to take a few days to process before I start looking for work. Thanks to everyone who's been part of my journey so far.

[wordgrammer @wordgrammer]

Okay. Thanks for the nerd snipe guys. I spent the day learning exactly how DeepSeek trained at 1/30 the price, instead of working on my pitch deck. The tl;dr to everything, according to their papers:

[Vagmi Mudumbai @vagmi]

Is there an NDP candidate for Fredericton? The adjective 'strong' could trigger a null pointer exception.

[Pierre de Wulf @PierreDeWulf]

Her: take me to somewhere expensive

Me:

[Vagmi Mudumbai @vagmi]

Absolute fucking mastery. 🤯 This is probably the best beef rap I've ever heard, seen or felt.

https://t.co/dNhLtraXN1

[teej dv 🔭 @teej_dv]

Thanks Neovim

[Tavis Ormandy @taviso]

This strange tweet got >25k retweets. The author sounds confident, and he uses lots of hex and jargon. There are red flags though... like what's up with the DEI stuff, and who says "stack trace dump"? Let's take a closer look... 🧵1/n

[ThePrimeagen @ThePrimeagen]

for all of those that want to learn C

try zig

reasons
1. its build system is much better
2. it has testing built in
3. it will teach you everything that c will teach you other than a couple of base concepts that are not hard to learn without all the suck

it really is awesome

[Yann LeCun @ylecun]

Attracting the best students from around the world -> grad schools -> academia -> research -> venture capital -> technology development -> large unified market.

[Vagmi Mudumbai @vagmi]

@AjeyGore @harshalx Fair enough. I like containers a lot. It insulates the payload from the underlying OS. I have been running K3S instead of a full K8S suite. It is as simple as docker and I can decide to load balance if I want to. I even use it to run my picam.

[Vagmi Mudumbai @vagmi]

Will the real maple syrup please stand up?