@varundhanwantri.bsky.social
Some of you know I'm terrible with names. Last year, I built a personal AI agent on my own VPS (because trust matters). Meet Kelly. She's on WhatsApp. Kelly handles voice notes and text.
Kelly remembers important events and people for me.
A year later? I cannot live without her.
Kelly remembers important events and people for me.
A year later? I cannot live without her.
February 2, 2026 at 4:13 PM
Some of you know I'm terrible with names. Last year, I built a personal AI agent on my own VPS (because trust matters). Meet Kelly. She's on WhatsApp. Kelly handles voice notes and text.
Kelly remembers important events and people for me.
A year later? I cannot live without her.
Kelly remembers important events and people for me.
A year later? I cannot live without her.
Not only OpenAI asks for IDs but the recent breach showed they shared names, email and location with an analytics company. When and why do we need personal info for analytics? Yet another reason to stay away from OpenAI.
November 27, 2025 at 4:13 PM
Not only OpenAI asks for IDs but the recent breach showed they shared names, email and location with an analytics company. When and why do we need personal info for analytics? Yet another reason to stay away from OpenAI.
Spent a learning & perfecting api design, system architecture, auth, deployment , cicd, devsecops, but never thought John Ousterhout's talk @Google would have a greater impact on my work! www.youtube.com/watch?v=bmSA...
A Philosophy of Software Design | John Ousterhout | Talks at Google
John Ousterhout, Professor of Computer Science at Stanford University, discusses complex techniques on how to become a more confident coder. John is excited to announce that he just published the…
www.youtube.com
November 25, 2025 at 8:50 AM
Spent a learning & perfecting api design, system architecture, auth, deployment , cicd, devsecops, but never thought John Ousterhout's talk @Google would have a greater impact on my work! www.youtube.com/watch?v=bmSA...
Wonder if I can create a custom post compiler hook for go and get better error results, hmm... might build a toy demo for myself
November 20, 2025 at 4:13 PM
Wonder if I can create a custom post compiler hook for go and get better error results, hmm... might build a toy demo for myself
Vibecoding doesn't help with Network Time Protocol, custom signature validations, complex server-client cryptography, docker deployment with python gcc vs clang builds, and much much more! Great tool to learn with tho...
November 20, 2025 at 8:50 AM
Vibecoding doesn't help with Network Time Protocol, custom signature validations, complex server-client cryptography, docker deployment with python gcc vs clang builds, and much much more! Great tool to learn with tho...
As some with a design background, I just realized, error messages are the UI of a compiler. See Rust compiler errors vs Go compiler errors.
November 8, 2025 at 8:50 AM
As some with a design background, I just realized, error messages are the UI of a compiler. See Rust compiler errors vs Go compiler errors.
I like how some LLMs end up with DROP TABLE recommendation, forgetting the 1000+ data rows when it faces a complex issue.
September 15, 2025 at 7:50 AM
I like how some LLMs end up with DROP TABLE recommendation, forgetting the 1000+ data rows when it faces a complex issue.
@alice-i-cecile.bsky.social inspirational work on Bevy! Looking forward to playing with it.
August 3, 2025 at 2:49 PM
@alice-i-cecile.bsky.social inspirational work on Bevy! Looking forward to playing with it.
Since working with Claude Code I’ve realized reviewing code better and faster is essential. Installed Vim bindings on VS Code and ended up down the NeoVim rabbit hole …
July 15, 2025 at 2:51 PM
Since working with Claude Code I’ve realized reviewing code better and faster is essential. Installed Vim bindings on VS Code and ended up down the NeoVim rabbit hole …
Glad to be part of the *winning team* at TechTO!
Special thanks to @elevenlabs.io and UnifiedTO for the amazing awards!
www.youtube.com/watch?v=Z6jh...
Special thanks to @elevenlabs.io and UnifiedTO for the amazing awards!
www.youtube.com/watch?v=Z6jh...
TechTO Hackathon 2025: PresenterThread: Transform Q&A Sessions
YouTube video by Vee D
www.youtube.com
June 30, 2025 at 11:25 PM
Glad to be part of the *winning team* at TechTO!
Special thanks to @elevenlabs.io and UnifiedTO for the amazing awards!
www.youtube.com/watch?v=Z6jh...
Special thanks to @elevenlabs.io and UnifiedTO for the amazing awards!
www.youtube.com/watch?v=Z6jh...
Claude Artifacts is out!
Create apps (artifacts) using Claude API from within Claude chat. Example artifacts are fun to play with but some are buggy.
Create apps (artifacts) using Claude API from within Claude chat. Example artifacts are fun to play with but some are buggy.
June 27, 2025 at 7:50 AM
Claude Artifacts is out!
Create apps (artifacts) using Claude API from within Claude chat. Example artifacts are fun to play with but some are buggy.
Create apps (artifacts) using Claude API from within Claude chat. Example artifacts are fun to play with but some are buggy.
Middle Management will be the first to be replaced by AI. Imagine these AIPMs available 24/7 on calls for project questions, next steps, progress checks and yea even a bit of therapy.
May 24, 2025 at 7:50 AM
Middle Management will be the first to be replaced by AI. Imagine these AIPMs available 24/7 on calls for project questions, next steps, progress checks and yea even a bit of therapy.
I don't think we realize how Gen AI (LLM) helps with productivity. Two years ago, creating a FAQ based on a website or PDF file would have taken a day or two.
Now, it takes a few seconds. Hallucinations? I can feed the FAQ + PDF to another LLM to test and spend time proofreading
Now, it takes a few seconds. Hallucinations? I can feed the FAQ + PDF to another LLM to test and spend time proofreading
May 2, 2025 at 7:50 AM
I don't think we realize how Gen AI (LLM) helps with productivity. Two years ago, creating a FAQ based on a website or PDF file would have taken a day or two.
Now, it takes a few seconds. Hallucinations? I can feed the FAQ + PDF to another LLM to test and spend time proofreading
Now, it takes a few seconds. Hallucinations? I can feed the FAQ + PDF to another LLM to test and spend time proofreading
⚠️ Every business owner must ask their web developers to implement llms.txt for their website! It helps ChatGPT, Groq, Perplexity, etc understand your website content. If a competitor has a cleanly written llms.txt, they would get priority over your website.
The /llms.txt file – llms-txt
A proposal to standardise on using an /llms.txt file to provide information to help LLMs use a website at inference time.
llmstxt.org
April 22, 2025 at 7:50 AM
⚠️ Every business owner must ask their web developers to implement llms.txt for their website! It helps ChatGPT, Groq, Perplexity, etc understand your website content. If a competitor has a cleanly written llms.txt, they would get priority over your website.
So far, I like writing my API servers in Python. But learning Golang will be crucial for Generative AI applications.
April 18, 2025 at 3:13 PM
So far, I like writing my API servers in Python. But learning Golang will be crucial for Generative AI applications.
"Vibe Coding" is like trying to get a toilet flush fixed without the plumber accidently remodling your kitchen. Sometimes it's several plumbers, and suddenly you have a hot tub in your bedroom. Yes thank you, I will keep the hot tub.
Great, my plumber left. I ran out of 💰
Great, my plumber left. I ran out of 💰
April 18, 2025 at 7:50 AM
"Vibe Coding" is like trying to get a toilet flush fixed without the plumber accidently remodling your kitchen. Sometimes it's several plumbers, and suddenly you have a hot tub in your bedroom. Yes thank you, I will keep the hot tub.
Great, my plumber left. I ran out of 💰
Great, my plumber left. I ran out of 💰
Finally, OpenAI released their Evals API. Earlier, I had to pitch my agents against each other to test against criteria. For example, have Sonya (QA AI Agent) call up Kelly AI for testing common questions. That eval was patched together with sticks and tape.
platform.openai.com/docs/guides/...
platform.openai.com/docs/guides/...
platform.openai.com
April 17, 2025 at 3:13 PM
Finally, OpenAI released their Evals API. Earlier, I had to pitch my agents against each other to test against criteria. For example, have Sonya (QA AI Agent) call up Kelly AI for testing common questions. That eval was patched together with sticks and tape.
platform.openai.com/docs/guides/...
platform.openai.com/docs/guides/...
Everyone is building AI Agent Platforms. Why do I need to bang my head against the wall "building an agent". Why can't I HIRE an agent as an employee?
This is what we are focusing on at Nameless Collective
This is what we are focusing on at Nameless Collective
April 14, 2025 at 3:13 PM
Everyone is building AI Agent Platforms. Why do I need to bang my head against the wall "building an agent". Why can't I HIRE an agent as an employee?
This is what we are focusing on at Nameless Collective
This is what we are focusing on at Nameless Collective
Funny how majority of executives think they need data scientists and machine learning experts to help create Generative AI applications.
April 13, 2025 at 3:13 PM
Funny how majority of executives think they need data scientists and machine learning experts to help create Generative AI applications.
We are moving from design systems to component registries. Some projects helping with this are uipub.com and @shadcn.
the shadcn/ui registry | UI.pub
Premium shadcn/ui components registry. UI Kits, Blocks, Templates.
uipub.com
April 12, 2025 at 7:50 AM
We are moving from design systems to component registries. Some projects helping with this are uipub.com and @shadcn.
I keep hearing, "we though Machine Learning and AI would help us work less - automate the mundane". But we forget the human condition. Aversion to change, greed and a need to hoard wealth.
April 3, 2025 at 3:13 PM
I keep hearing, "we though Machine Learning and AI would help us work less - automate the mundane". But we forget the human condition. Aversion to change, greed and a need to hoard wealth.
I want my reasoning model to take it's own time! Let it work overnight, but slower - AND cheaper.
March 16, 2025 at 4:13 PM
I want my reasoning model to take it's own time! Let it work overnight, but slower - AND cheaper.
@groqinc.bsky.social Slow inference is not a bug. It's a feature.
Cheap, Reliable, Fast. Pick two? For some models I'll take Cheap and Reliable! Some of workflow can be done late at night when I'm asleep... Let it be a "good" but "slower" inference. PS: Glad you are expanding in Saudi.
Cheap, Reliable, Fast. Pick two? For some models I'll take Cheap and Reliable! Some of workflow can be done late at night when I'm asleep... Let it be a "good" but "slower" inference. PS: Glad you are expanding in Saudi.
January 23, 2025 at 4:24 PM
@groqinc.bsky.social Slow inference is not a bug. It's a feature.
Cheap, Reliable, Fast. Pick two? For some models I'll take Cheap and Reliable! Some of workflow can be done late at night when I'm asleep... Let it be a "good" but "slower" inference. PS: Glad you are expanding in Saudi.
Cheap, Reliable, Fast. Pick two? For some models I'll take Cheap and Reliable! Some of workflow can be done late at night when I'm asleep... Let it be a "good" but "slower" inference. PS: Glad you are expanding in Saudi.
Ok, I think my agent Kelly is coming along well. She's aware of what I am working on, my schedule and and can send a link to book a meeting with me. All on WhatsApp.
January 22, 2025 at 8:50 AM
Ok, I think my agent Kelly is coming along well. She's aware of what I am working on, my schedule and and can send a link to book a meeting with me. All on WhatsApp.
Creating and managing two personal agents is challenging today! One on ElevenLabs for voice and another for WhatsApp. Both share persona, context and data. Voice and text persona's (system message) have different formatting requirements.
I wonder if there's a way to centralize?
I wonder if there's a way to centralize?
Persona – Centrul de sănătate și frumusețe
persona.md
January 20, 2025 at 8:50 AM
Creating and managing two personal agents is challenging today! One on ElevenLabs for voice and another for WhatsApp. Both share persona, context and data. Voice and text persona's (system message) have different formatting requirements.
I wonder if there's a way to centralize?
I wonder if there's a way to centralize?