iiinova — where your digital self keeps evolving

Hardik @hardikvala.com · 10/01/2026

tried some hyperlocal queries against my fav ai web searches, e.g. - "Massage parlors that offer couples massages in Soma, SF" - "All cafes open past 6 pm PT on Sunday in downtown San Mateo, with space to work and outlets." disappointed by the results 😞

000

Hardik @hardikvala.com · 09/01/2026

when you accidentally leave your crawler running overnight 🤦‍♂️

000

Hardik @hardikvala.com · 23/12/2025

Yelp's AI search: Rating: 1 star Review: I asked, "give me cafes that are open past 6 pm PT on a sunday in downtown san mateo, with space to work and outlets." Only 1 result is in san mateo, and it's not located downtown.

010

Hardik @hardikvala.com · 19/12/2025

If you're not using the most powerful model for your therapist prompt, you don't love yourself.

010

Hardik @hardikvala.com · 16/12/2025

founder Sanity = traction − comparison...boy, I'm deep in the negative right now.

000

Hardik @hardikvala.com · 29/11/2025

man, the #claude ux is not friendly to mcp connectors serving images. my bit.ly/tensordiagram mcp server can add tensor visuals to help with deep learning explainers, but the end ux here is not seamless...

040

Hardik @hardikvala.com · 28/11/2025

nobody asked for somebody to create a a visual explainer for torch.linspace. I did it anyways. #TensorViz

030

Hardik @hardikvala.com · 27/11/2025

burned too many tokens to get CC to draw a tensor with thanksgiving colors using bit.ly/tensordiagram

000

Hardik @hardikvala.com · 26/11/2025

I came across this meme over a year ago, and it still makes me laugh.

000

Hardik @hardikvala.com · 25/11/2025

Got addicted to @srushnlp.bsky.social 's Tensor Puzzles, so I wrote a sequel with more puzzles: github.com/hardik-vala/.... Example:

000

Hardik @hardikvala.com · 21/11/2025

using None as an index value for torch tensors adds a new dim. ludicrously simple when visualized.

000

Hardik @hardikvala.com · 20/11/2025

finally made sense of all the ways you can index tensors in pytorch / numpy #TensorViz

000

Hardik @hardikvala.com · 17/11/2025

"Quit the wrong stuff. Stick with the right stuff. Have the guts to do one or the other." — Seth Godin, The Dip A questionnaire I've developed up over years to navigate major stick or quit decisions: hardikvala.com/blog/stick-o... Answer with your gut, not your brain.:

hardikvala.com

⛰️ stick or quit: a manual

hardik vala: technical founder (yc s23, ex-google, ex-apple)

100

Hardik @hardikvala.com · 13/11/2025

Creating these visuals is addictive, but getting out of hand. Had to see the difference between torch chunk and unbind, but this visual doesn't need to exist. #TensorViz

000

Hardik @hardikvala.com · 12/11/2025

Not my best work. Created a visual of torch's transpose op over 3d tensors, but I don't know if the random coloring is instructive or confusing. #TensorViz

100

Hardik @hardikvala.com · 12/11/2025

TIL `tensor[None]` is just a shorthand for `tensor.unsqueeze( 0)` in torch.

000

Hardik @hardikvala.com · 11/11/2025

seriously!? every time i open the app

000

Hardik @hardikvala.com · 10/11/2025

Me: “who’s your favorite entrepreneur?” GF: “you” Me: “I’m not an entrepreneur. I’m a fraud.” GF: “that’s most entrepreneurs” Me: “good point”

010

Hardik @hardikvala.com · 10/11/2025

Visual explainer for the pytorch op, `torch.unbind` #TensorViz

000

Hardik @hardikvala.com · 09/11/2025

From Seth Godin's book, The Dip.

000

Hardik @hardikvala.com · 07/11/2025

This is a cool resource: github.com/nayafia/lemo... . Documents a bunch of ways to make money with open source, with pros and cons of each approach.

github.com

GitHub - nayafia/lemonade-stand: A handy guide to financial support for open source

A handy guide to financial support for open source - nayafia/lemonade-stand

000

Hardik @hardikvala.com · 07/11/2025

Finished the visual explainer for the torch op, `chunk`. #TensorViz

000

Hardik @hardikvala.com · 06/11/2025

Started a visual explainer for the torch op, `chunk`. The full explainer will have different arg combinations for no. of chunks and dim.

000

Hardik @hardikvala.com · 05/11/2025

Framework lock-in is definitely not a day 0 / pre-PMF concern. mlechner.substack.com/p/why-we-sta...

000

Hardik @hardikvala.com · 05/11/2025

whytorch.org is quietly addictive. It's been fun playing with pytorch tensors in a visual format. Not affiliated with the project. Just a fan.

000

Hardik @hardikvala.com · 02/11/2025

torch.outer(a, b), visualized #TensorViz (drawn w/ bit.ly/tensordiagram)

000

Hardik @hardikvala.com · 02/11/2025

Any time you see t[:, None] on a 1d tensor, it's just transpose #TensorViz

000

Hardik @hardikvala.com · 31/10/2025

if you, like me, forget how tensors get reshaped. (drawn with bit.ly/tensordiagram)

000

Hardik @hardikvala.com · 30/10/2025

Fed up with trying to imagine chains of tensor transformations in my head, wrote a library to relieve my pain: github.com/hardik-vala/.... Lets you visualize torch, jax, tf, numpy tensors in notebooks and other python contexts, for understanding and debugging. Enjoy.

010

Hardik @hardikvala.com · 29/10/2025

is playground mode a new feature in colab? liking it for messing around with a notebook you authored that's also shared

000

Hardik @hardikvala.com · 29/10/2025

JAX / TPU docs are incomplete. For anyone looking to run Pallas kernels (JAX's kernel lang) on TPUs, make sure you're running Python 3.10+.

000

Hardik @hardikvala.com · 29/10/2025

Tensor Logic does not seem all that different from torch and tf, and don't see how it solves hallucinations. The unification of symbolic logic, database operations, and tensor transformations was intellectually instructive tho. arxiv.org/abs/2510.12269

arxiv.org

Tensor Logic: The Language of AI

Progress in AI is hindered by the lack of a programming language with all the requisite features. Libraries like PyTorch and TensorFlow provide automatic differentiation and efficient GPU implementati...

000

Hardik @hardikvala.com · 28/10/2025

When you create a typo in the process of fixing one. Entropy refuses to go down.

000

Hardik @hardikvala.com · 18/10/2025

A remix on llama2.c: ModernBERT, in one file of pure C: github.com/hardik-vala/...

github.com

000

Hardik @hardikvala.com · 13/10/2025

How do people visualize tensors? especially when reading foreign pytorch / jax code.

000

Hardik @hardikvala.com · 03/10/2025

Takeaways from the Flower AI conference last week on decentralized / federated AI: • 𝗦𝗵𝗶𝗳𝘁 𝗶𝗻 𝗯𝗿𝗮𝗻𝗱𝗶𝗻𝗴 from "federated" and "privacy", to "decentralized" and "open".

100

Hardik @hardikvala.com · 25/09/2025

Took a break from CC, and took Gemini CLI for a spin. Good DX still lacking: * No indicator in terminal telling you # code lines highlighted in IDE * Tends to make spurious formatting changes * Gets stuck in tool calling loops

000

Hardik @hardikvala.com · 25/09/2025

One day, I would like my name to be immortalized as an LLM vocab token.

100

Hardik @hardikvala.com · 24/09/2025

Find myself frequently saying, "Back when it was called ML...".

010

Hardik @hardikvala.com · 17/09/2025

omg, the ONNX runtime is wicked fast. ran some perf benchmarks for ModernBERT on cpu vs gpu (rtx 4070) and ONNX was 3x faster than vanilla torch. not superior on cpu tho.

010

Hardik @hardikvala.com · 17/09/2025

So any model trained/fine-tuned with llama outputs — even if it's not based on llama's architecture at all — must be prefixed with "llama" 😲 . e.g. huggingface.co/ai4privacy/l... (ModernBert fine-tune) License: www.llama.com/llama3_1/lic...

000

Hardik @hardikvala.com · 16/09/2025

It's funny how the accessibility features in browsers make websites more accessible for browser agents.

000

Hardik @hardikvala.com · 16/09/2025

i thought including the onloading/offloading of tensors to gpu mem would add significant lag when running batch llm inference, but it turns out those ops are negligible. makes sense when your gpu's mem bandwidth is 1450 gb/s. throughput: w/ mem ops: 155.7 tok/s w/o mem ops: 156.1 tok/s

100

Hardik @hardikvala.com · 13/09/2025

I can't decide if #CompoundAI is fatuous term that refers to something obvious. It's like marketing cars as Compound Engine. Of course AI is integrated with other components to build complete applications.

000

Hardik @hardikvala.com · 13/09/2025

OpenAI's hallucination paper is gaining attention, but in case you missed it FAIR published AbstentionBench a few mos ago: Benchmark for measuring LLM capacity to *abstain* against questions with underspecification, false premises, subjective interpretations, etc. arxiv.org/abs/2506.090...

000

Hardik @hardikvala.com · 12/09/2025

PyTorch installs have unseated code compilations.

000

Hardik @hardikvala.com · 12/09/2025

Thanks @mlopscommunity.bsky.social for the shoutout in today's newsletter! www.controlpla.in/blog/prompt-...

011

Hardik @hardikvala.com · 11/09/2025

minusx.ai/blog/decodin... - Takeaways: * single-agent > multi-agent * small models > large models (for low entropy tasks) * llm search > RAG * PLEASE, NEVER, ALWAYS, IMPORTANT, etc. are still the best ways to steer LLM behavior

minusx.ai

What makes Claude Code so damn good (and how to recreate that magic in your agent)!?

Claude Code is the most delightful AI agent/workflow I have used so far. Not only does it make targeted edits or vibe coding throwaway tools less annoying, ...

000