Ashish
@tulsani.net
Digital, Data, analytics, insights, AI, Agents, LLMs, Memes
tulsani.net
tulsani.net
Pinned
we might see LLM-powered systems like Grok OS, Claude OS, and more changing the way how devices work.
August 22, 2025 at 2:33 AM
we might see LLM-powered systems like Grok OS, Claude OS, and more changing the way how devices work.
Testing for a post. It's interesting :)
August 14, 2025 at 6:09 PM
Testing for a post. It's interesting :)
For anyone building, deploying, or auditing language models, integrating persona vector methods represents an actionable best practice.
open.substack.com/pub/tulsani/...
open.substack.com/pub/tulsani/...
Personality Traits in Large Language Models
Monitoring, Predicting, and Controlling Behavioral Shifts via Persona Vectors
open.substack.com
August 14, 2025 at 3:43 AM
For anyone building, deploying, or auditing language models, integrating persona vector methods represents an actionable best practice.
open.substack.com/pub/tulsani/...
open.substack.com/pub/tulsani/...
I built a ChatGPT prompt that tests any startup idea through the lens of 100 legendary thinkers.
It shows you what to build, what to avoid, and why.
Want it? I will be sharing it with my subscribers!
It shows you what to build, what to avoid, and why.
Want it? I will be sharing it with my subscribers!
August 12, 2025 at 8:39 PM
I built a ChatGPT prompt that tests any startup idea through the lens of 100 legendary thinkers.
It shows you what to build, what to avoid, and why.
Want it? I will be sharing it with my subscribers!
It shows you what to build, what to avoid, and why.
Want it? I will be sharing it with my subscribers!
You do not need AGI. You need a cache!
August 9, 2025 at 3:13 AM
You do not need AGI. You need a cache!
Every model release: Cheaper, Faster, Smarter!
My Laptop: Louder, Hotter, Braver!
My Laptop: Louder, Hotter, Braver!
August 8, 2025 at 1:16 PM
Every model release: Cheaper, Faster, Smarter!
My Laptop: Louder, Hotter, Braver!
My Laptop: Louder, Hotter, Braver!
Testing persona vectors as described in Anthropic’s latest paper: How to Monitor and control character traits in language models. I ran an experiment with intentional hallucination to see how well trait steering works. AI safety isn’t just theory, it’s hands-on!
August 3, 2025 at 2:35 AM
Testing persona vectors as described in Anthropic’s latest paper: How to Monitor and control character traits in language models. I ran an experiment with intentional hallucination to see how well trait steering works. AI safety isn’t just theory, it’s hands-on!
Testing persona vectors as described in Anthropic’s latest paper: How to Monitor and control character traits in language models. I ran an experiment with intentional hallucination to see how well trait steering works. AI safety isn’t just theory, it’s hands-on!
August 2, 2025 at 4:25 PM
Testing persona vectors as described in Anthropic’s latest paper: How to Monitor and control character traits in language models. I ran an experiment with intentional hallucination to see how well trait steering works. AI safety isn’t just theory, it’s hands-on!
Reposted by Ashish
Please don't plug keyboards or other devices into random USB ports, ok?
July 29, 2025 at 10:47 AM
Please don't plug keyboards or other devices into random USB ports, ok?
Got an invite to use @PerplexityComet , I will be trying out few interesting things!
July 28, 2025 at 5:00 PM
Got an invite to use @PerplexityComet , I will be trying out few interesting things!
If you’re building an AI agent, I’d love to volunteer to test it for ethics, trust, bias, and responsible AI behavior. I’m offering a free audit report based on a comprehensive testing framework. Reach out if you’re interested.
July 27, 2025 at 8:39 PM
If you’re building an AI agent, I’d love to volunteer to test it for ethics, trust, bias, and responsible AI behavior. I’m offering a free audit report based on a comprehensive testing framework. Reach out if you’re interested.
How do you manage everything when there’s so much to read, build, learn, ship, explore, write, and grow and all at once?
July 24, 2025 at 7:24 PM
How do you manage everything when there’s so much to read, build, learn, ship, explore, write, and grow and all at once?
Pro tip: Always use ordered list in emails, easy to refer!
July 24, 2025 at 7:24 AM
Pro tip: Always use ordered list in emails, easy to refer!
70% dark Lindt are my new best friend!
July 23, 2025 at 10:05 AM
70% dark Lindt are my new best friend!
Anyone know any ai based solution where i can drop voice notes and texts and the agent performs tasks like organizing the notes as per relevant topic, calendar, email and to do lists.
July 22, 2025 at 1:55 PM
Anyone know any ai based solution where i can drop voice notes and texts and the agent performs tasks like organizing the notes as per relevant topic, calendar, email and to do lists.
The notion of causality arises from our discomfort with the unfamiliar and our tendency to seek the familiar in the new.
Predictability comes not from laws or necessity, but from repeated patterns. Regular occurrence does not imply a deterministic cause.
Predictability comes not from laws or necessity, but from repeated patterns. Regular occurrence does not imply a deterministic cause.
July 20, 2025 at 5:25 PM
The notion of causality arises from our discomfort with the unfamiliar and our tendency to seek the familiar in the new.
Predictability comes not from laws or necessity, but from repeated patterns. Regular occurrence does not imply a deterministic cause.
Predictability comes not from laws or necessity, but from repeated patterns. Regular occurrence does not imply a deterministic cause.
Agents without guardrails are like race cars without brakes. Fun - until they crash.
open.substack.com/pub/tulsani/...
open.substack.com/pub/tulsani/...
Why Every AI Agent Needs Guardrails
How to protect your AI workflows with layered safety checks
open.substack.com
July 18, 2025 at 12:20 PM
Agents without guardrails are like race cars without brakes. Fun - until they crash.
open.substack.com/pub/tulsani/...
open.substack.com/pub/tulsani/...
Agents can interpret natural language and operate tools. But what happens when the input is misleading, manipulative, or flat-out malicious?
open.substack.com/pub/tulsani/...
open.substack.com/pub/tulsani/...
Why Every AI Agent Needs Guardrails
How to protect your AI workflows with layered safety checks
open.substack.com
July 18, 2025 at 12:18 PM
Agents can interpret natural language and operate tools. But what happens when the input is misleading, manipulative, or flat-out malicious?
open.substack.com/pub/tulsani/...
open.substack.com/pub/tulsani/...
Grok introduced me to Mermaid Chart, and I can't stop using it!
July 17, 2025 at 3:30 AM
Grok introduced me to Mermaid Chart, and I can't stop using it!
A brilliant example of applied LLMs in low-resource, high-impact settings.
Dr.Copilot is a multi-agent assistant that helps Romanian doctors improve how they communicate with patients in telemedicine. It focuses only on the quality of the message, not the medical content.
Dr.Copilot is a multi-agent assistant that helps Romanian doctors improve how they communicate with patients in telemedicine. It focuses only on the quality of the message, not the medical content.
July 16, 2025 at 8:07 PM
A brilliant example of applied LLMs in low-resource, high-impact settings.
Dr.Copilot is a multi-agent assistant that helps Romanian doctors improve how they communicate with patients in telemedicine. It focuses only on the quality of the message, not the medical content.
Dr.Copilot is a multi-agent assistant that helps Romanian doctors improve how they communicate with patients in telemedicine. It focuses only on the quality of the message, not the medical content.
Turned my iPhone to grayscale. It sucks out the soul but is good for reading and fewer distractions. De-colouring discourages fiddling and doomscrolling. Highly Recommend!
July 16, 2025 at 6:35 PM
Turned my iPhone to grayscale. It sucks out the soul but is good for reading and fewer distractions. De-colouring discourages fiddling and doomscrolling. Highly Recommend!
Claude connectors looks interesting!
July 16, 2025 at 5:10 PM
Claude connectors looks interesting!