Sai Prakash
@sylonzero.bsky.social
CTO & Cofounder, MindStaq • mindstaq.com • Product Design + UX + Engineering • Brooklyn, by way of India • formerly Microsoft, Reuters, Agility • Blog: robotmunki.com
Pinned
Sai Prakash
@sylonzero.bsky.social
· Aug 24
I spend most of my day in either coding, running a startup or music (I play the tabla) + photography.
Here's what I mostly write about in this account:
- Code / AI-related experiments, tools
- Security incidents that alarm me
- Engineering architecture, algorithms
- A few photos from time to time
Here's what I mostly write about in this account:
- Code / AI-related experiments, tools
- Security incidents that alarm me
- Engineering architecture, algorithms
- A few photos from time to time
Useless but entertaining fact of the day - have a great weekend all! troypress.com/1973-impleme...
1973 Implementation of Wordle was Published by DEC
While some have traced Wordle to Lingo, a game show that started in 1987, they’ve missed an earlier implementation: WORD was published in 101 Computer Games by Digital Equipment Corp. in 1973 and l…
troypress.com
November 7, 2025 at 7:23 PM
Useless but entertaining fact of the day - have a great weekend all! troypress.com/1973-impleme...
Reposted by Sai Prakash
My new field guide to alternatives to standard LLMs:
Gated DeltaNet hybrids (Qwen3-Next, Kimi Linear), text diffusion, code world models, and small reasoning transformers.
🔗 magazine.sebastianraschka.com/p/beyond-sta...
Gated DeltaNet hybrids (Qwen3-Next, Kimi Linear), text diffusion, code world models, and small reasoning transformers.
🔗 magazine.sebastianraschka.com/p/beyond-sta...
November 4, 2025 at 2:49 PM
My new field guide to alternatives to standard LLMs:
Gated DeltaNet hybrids (Qwen3-Next, Kimi Linear), text diffusion, code world models, and small reasoning transformers.
🔗 magazine.sebastianraschka.com/p/beyond-sta...
Gated DeltaNet hybrids (Qwen3-Next, Kimi Linear), text diffusion, code world models, and small reasoning transformers.
🔗 magazine.sebastianraschka.com/p/beyond-sta...
Gorgeous space photo of the day: the Red Spider Nebula
www.planetary.org/space-images...
www.planetary.org/space-images...
The Red Spider Nebula
The Red Spider Nebula, also known as NGC 6537, as imaged in the infrared by the NASA/ESA/CSA James Webb Space Telescope.
www.planetary.org
November 4, 2025 at 10:35 PM
Gorgeous space photo of the day: the Red Spider Nebula
www.planetary.org/space-images...
www.planetary.org/space-images...
Haven't given this a whirl yet, but the idea of a vulnerability scanner for your MCP services is a good one of course: github.com/cisco-ai-def... #ai #security #mcp
GitHub - cisco-ai-defense/mcp-scanner: Scan MCP Servers for vulnerabilities
Scan MCP Servers for vulnerabilities. Contribute to cisco-ai-defense/mcp-scanner development by creating an account on GitHub.
github.com
October 28, 2025 at 3:43 AM
Haven't given this a whirl yet, but the idea of a vulnerability scanner for your MCP services is a good one of course: github.com/cisco-ai-def... #ai #security #mcp
Always amazing to read stories about the greats like Ken Thompson but this para at the end after he joined Google: "But eventually Thompson joined the three-person team that would create the programming language Go." 🤯 Just his latest contribution...
Full article: thenewstack.io/ken-thompson...
Full article: thenewstack.io/ken-thompson...
Ken Thompson Recalls Unix's Rowdy, Lock-Picking Origins
Ken Thompson's vivid recollection of the rowdy roomful of geeks at Bell Labs who built the digital world in a spirit of open play.
thenewstack.io
October 27, 2025 at 5:12 PM
Always amazing to read stories about the greats like Ken Thompson but this para at the end after he joined Google: "But eventually Thompson joined the three-person team that would create the programming language Go." 🤯 Just his latest contribution...
Full article: thenewstack.io/ken-thompson...
Full article: thenewstack.io/ken-thompson...
Reposted by Sai Prakash
I've been using @tanstack.com Start for a new project and it's super good. The server functions completely replace the need for TRPC/GraphQL/REST, the middleware is composable and fully typed, and having TSRouter's nice typing and stateful search params is icing on the cake. A+!
October 25, 2025 at 11:30 AM
I've been using @tanstack.com Start for a new project and it's super good. The server functions completely replace the need for TRPC/GraphQL/REST, the middleware is composable and fully typed, and having TSRouter's nice typing and stateful search params is icing on the cake. A+!
Always amazed at the work in PyTorch - it's already a backbone for so much ML work - and now they release: Monarch, a distributed programming framework that brings the simplicity of single-machine PyTorch to entire clusters
pytorch.org/blog/introdu...
pytorch.org/blog/introdu...
Introducing PyTorch Monarch – PyTorch
pytorch.org
October 23, 2025 at 12:13 PM
Always amazed at the work in PyTorch - it's already a backbone for so much ML work - and now they release: Monarch, a distributed programming framework that brings the simplicity of single-machine PyTorch to entire clusters
pytorch.org/blog/introdu...
pytorch.org/blog/introdu...
Reposted by Sai Prakash
I recorded a ten minute video showing my vibe-coding process for building a tool for sharing formatted terminal sessions via copy and paste using the new Claude Code for web - now available on YouTube here www.youtube.com/watch?v=GQvM...
More notes on my blog: simonwillison.net/2025/Oct/23/...
More notes on my blog: simonwillison.net/2025/Oct/23/...
Using Claude Code for web to build a tool to copy-paste share terminal sessions
YouTube video by Simon Willison
www.youtube.com
October 23, 2025 at 4:17 AM
I recorded a ten minute video showing my vibe-coding process for building a tool for sharing formatted terminal sessions via copy and paste using the new Claude Code for web - now available on YouTube here www.youtube.com/watch?v=GQvM...
More notes on my blog: simonwillison.net/2025/Oct/23/...
More notes on my blog: simonwillison.net/2025/Oct/23/...
Right after I posted my rant on VZero enshittification, I see this pop up on Wired lol: www.wired.com/story/can-ai...
Can AI Avoid the Enshittification Trap?
Cory Doctorow’s theory of “enshittification” explains how tech platforms rot from within. As AI grows more profitable—and powerful—it risks the same fate.
www.wired.com
October 21, 2025 at 3:14 PM
Right after I posted my rant on VZero enshittification, I see this pop up on Wired lol: www.wired.com/story/can-ai...
@vercel.com has really disappointed me with the latest V0 platform. Trusting that they will uphold quality has become a frequent source of disappointment with AI-based apps these days. Poor model choices, loss of a previous fine-tuned model that had amazing output, bad UI/UX. I am so bummed. #vercel
October 16, 2025 at 10:32 PM
@vercel.com has really disappointed me with the latest V0 platform. Trusting that they will uphold quality has become a frequent source of disappointment with AI-based apps these days. Poor model choices, loss of a previous fine-tuned model that had amazing output, bad UI/UX. I am so bummed. #vercel
@vercel.community @v0.dev What happened to v0? The new designs coming out of the chat for basic things look terrible. The model being used is making repeated mistakes where previously it was perfect! Something has gone terribly wrong with the platform update...is it the switch to GPT5? Tailwind 4?
October 16, 2025 at 10:23 PM
@vercel.community @v0.dev What happened to v0? The new designs coming out of the chat for basic things look terrible. The model being used is making repeated mistakes where previously it was perfect! Something has gone terribly wrong with the platform update...is it the switch to GPT5? Tailwind 4?
Claude Skills - it's interesting how AI features basically mirror things that human devs do as we mature and get more organized. Considering how well-known this is, I'm surprised that this stuff doesn't come out faster from the AI platforms out there... www.anthropic.com/news/skills
October 16, 2025 at 7:54 PM
Claude Skills - it's interesting how AI features basically mirror things that human devs do as we mature and get more organized. Considering how well-known this is, I'm surprised that this stuff doesn't come out faster from the AI platforms out there... www.anthropic.com/news/skills
Reposted by Sai Prakash
Chapter 3, and with it the first 176 pages, is now live! (mng.bz/lZ5B)
October 16, 2025 at 1:35 PM
Chapter 3, and with it the first 176 pages, is now live! (mng.bz/lZ5B)
Vermont vibes this morning #smugglersnotch
October 12, 2025 at 6:15 PM
Vermont vibes this morning #smugglersnotch
Definitely an ecosystem one cannot ignore, due to the sheer opportunity volume: developers.openai.com/apps-sdk #openai #chatgpt #apps
Apps SDK
Learn how to use Apps SDK by OpenAI. Our framework to build apps for ChatGPT.
developers.openai.com
October 7, 2025 at 1:22 AM
Definitely an ecosystem one cannot ignore, due to the sheer opportunity volume: developers.openai.com/apps-sdk #openai #chatgpt #apps
Reposted by Sai Prakash
If you've been trying to figure out DSPy - the automatic prompt optimization system - this talk by @dbreunig.bsky.social is the clearest explanation I've seen yet, with a very useful real-world case study www.youtube.com/watch?v=I9Zt...
My notes here: simonwillison.net/2025/Oct/4/d...
My notes here: simonwillison.net/2025/Oct/4/d...
Let the LLM Write the Prompts: An Intro to DSPy in Compound AI Pipelines
YouTube video by Databricks
www.youtube.com
October 4, 2025 at 11:05 PM
If you've been trying to figure out DSPy - the automatic prompt optimization system - this talk by @dbreunig.bsky.social is the clearest explanation I've seen yet, with a very useful real-world case study www.youtube.com/watch?v=I9Zt...
My notes here: simonwillison.net/2025/Oct/4/d...
My notes here: simonwillison.net/2025/Oct/4/d...
This is a very useful background on the architecture of learning systems: little-book-of.github.io/maths/books/... #ai #ml #machinelearning
Chapter 8. The Architecture of Learning: From Statistics to Intelligence – The Little Chronicles of Mathematics, and the Mind of Machines (TLC3M)
little-book-of.github.io
October 4, 2025 at 3:53 PM
This is a very useful background on the architecture of learning systems: little-book-of.github.io/maths/books/... #ai #ml #machinelearning
ChatGPT-5 is facing the issue of any chat tool that really tries to insert a ton of parallel routing/retrieval and other steps before responding to a request. I'd be curious to know what the time-to-first-token value is - as an end user, it's pretty slow and starting to annoy...
October 4, 2025 at 3:27 PM
ChatGPT-5 is facing the issue of any chat tool that really tries to insert a ton of parallel routing/retrieval and other steps before responding to a request. I'd be curious to know what the time-to-first-token value is - as an end user, it's pretty slow and starting to annoy...
The value of what you're creating might be directly proportional to how much you worry about the laptop on which it's happening. I've started to worry about my M4 MacBook Pro for the first time in a long time... #randomthoughts
October 3, 2025 at 3:46 PM
The value of what you're creating might be directly proportional to how much you worry about the laptop on which it's happening. I've started to worry about my M4 MacBook Pro for the first time in a long time... #randomthoughts
Is Claude really down? Getting a downtime warning on the desktop app…
October 3, 2025 at 2:15 AM
Is Claude really down? Getting a downtime warning on the desktop app…
Sometimes there's a bump in the road...slow down, deal with it and keep going.
#Brooklyn #bumpintheroad #Fujifilm #photography #cityscapes
#Brooklyn #bumpintheroad #Fujifilm #photography #cityscapes
October 2, 2025 at 8:13 PM
Sometimes there's a bump in the road...slow down, deal with it and keep going.
#Brooklyn #bumpintheroad #Fujifilm #photography #cityscapes
#Brooklyn #bumpintheroad #Fujifilm #photography #cityscapes
Getting closer to the point where we might actually have to dust off the math books and get cracking on figuring out quantum algorithms and programming techniques - Shor's algorithm is probably a good starting point...
blog.google/technology/r...
blog.google/technology/r...
We’re scaling quantum computing even faster with Atlantic Quantum.
Google Quantum AI was founded in 2012 and our mission today remains the same — build quantum computers that can solve otherwise unsolvable problems. We’re making steady …
blog.google
October 2, 2025 at 7:38 PM
Getting closer to the point where we might actually have to dust off the math books and get cracking on figuring out quantum algorithms and programming techniques - Shor's algorithm is probably a good starting point...
blog.google/technology/r...
blog.google/technology/r...
A poignant point of view on why the immutability of email matters in a world full of AI: www.fastmail.com/blog/not-wri...
This blog post was not written with AI
Email immutability matters even more in a world of AI. Here is how we enable your use of your data, in line with our values.
www.fastmail.com
October 2, 2025 at 7:32 PM
A poignant point of view on why the immutability of email matters in a world full of AI: www.fastmail.com/blog/not-wri...
Interesting API (and first product) for fine-tuning models from Thinking Machines: thinkingmachines.ai/blog/announc... #ai #llm #ml #machinelearning #Tinker
Announcing Tinker
Introducing Tinker: a flexible API for fine-tuning language models.
thinkingmachines.ai
October 2, 2025 at 1:16 PM
Interesting API (and first product) for fine-tuning models from Thinking Machines: thinkingmachines.ai/blog/announc... #ai #llm #ml #machinelearning #Tinker
📊 Key facts from this article shared by Meredith:
• Double video length → 4x more energy
• A 5-second clip ≈ 1 hour of microwave use
• AI already = ~20% of datacenter power
This raises tough questions:
• Can we innovate and stay energy-efficient?
• Should AI platforms report power metrics?
• Double video length → 4x more energy
• A 5-second clip ≈ 1 hour of microwave use
• AI already = ~20% of datacenter power
This raises tough questions:
• Can we innovate and stay energy-efficient?
• Should AI platforms report power metrics?
October 1, 2025 at 7:23 PM
📊 Key facts from this article shared by Meredith:
• Double video length → 4x more energy
• A 5-second clip ≈ 1 hour of microwave use
• AI already = ~20% of datacenter power
This raises tough questions:
• Can we innovate and stay energy-efficient?
• Should AI platforms report power metrics?
• Double video length → 4x more energy
• A 5-second clip ≈ 1 hour of microwave use
• AI already = ~20% of datacenter power
This raises tough questions:
• Can we innovate and stay energy-efficient?
• Should AI platforms report power metrics?