Vaibhav
@vaibhav_pandey
Curious about intelligence and reasoning. Building @tradestack_uk.
You might like
one machine manipulates knowledge faster the other manipulates information faster
only those abstractions win which people can't get bored or maintaining/improving
most enterprise AI startups are (unknowingly) building a backend utility kit for a new collaboration software for their vertical/function
Unpopular opinion - "jobs to be done" feels like an incomplete framework when thinking about AI products. Because the list of jobs and the success criteria for jobs keep getting rewritten everyday.
the difference between progress and stagnation seems to stem from whether you are acting on your on thoughts or acting on reality
Brilliant demo - first time I felt a sense of what an AI first OS for a function/vertical could actually be like. Clearly a team with in-depth customer x technical insights and execution ability.
We have raised a $61M Series A to automate customer operations. The world’s leading companies like DoorDash trust Giga to supercharge customer experience with AI.
Today I taught a PhD class on the economics of AI. In doing so, I drew this picture on the board of my current understanding of what I called the "how good is AI" literature (aka the productivity impacts of AI). I thought I'd write up a long version of that discussion - at the…
Was thinking about this idea for the past few months from the lens of speed and reliability
New on the Anthropic Engineering blog: tips on how to build more efficient agents that handle more tools while using fewer tokens. Code execution with the Model Context Protocol (MCP): anthropic.com/engineering/co…
Updating the mental model to: emergent systems cannot be controlled but given an opportunity to self-align - the strength of alignment is a key variable
Intuitively I think there are only two ways to control emergent systems - explainability or self-organisation.
note to self: information becomes knowledge when it enters an intelligent vessel
I'll take vibe coding platforms seriously when they pay me royalties for my specs rather than charging me for code generation and deployment.
ability to generalise is basically linked to the the design of the systems that generated data the model was trained on: Design > data distribution > training generalisation
Interesting read - Time to bring back "sleep time compute"? On a more serious note a more practical challenge is how to update your rest of the system in sync with this adapting context - particularly the tooling layer - unless that itself is implemented as a meta layer
New Stanford + SambaNova + UC Berkeley paper proposes quite a revolutionary idea. 🤯 Proves LLMs can be improved by purely changing the input context, instead of changing weights. Introduces a new method called Agentic Context Engineering (ACE). It helps language models…
Intuitively I think there are only two ways to control emergent systems - explainability or self-organisation.
UX design went from: “how do you layout this page so that people click BUY most often” to: “how do you channel the machine god’s intelligence through this meager, unworthy vessel”
Been using mem0 for over a year now and been seeing @taranjeetio ‘s persistence for about a decade. 🚀🚀
Memory is what makes us human. It's also what makes AI truly intelligent. @mem0ai has raised $24M to build the universal memory layer for AI. Thousands of teams in production. 14M downloads. 41K GitHub stars. Intelligence needs memory & we're building it for everyone. More👇
Found a very curious behaviour in claude sonnet 4.5 where for lengthy parameters in a tool call, the model just skips the parameter and can’t get it right despite any change in the instruction. Seems that the tool call is being modified during inference as the model has no clue.
United States Trends
- 1. Kevin James 5,623 posts
- 2. Bubba 35.8K posts
- 3. Bill Clinton 126K posts
- 4. Jack Hughes 2,290 posts
- 5. Metroid 12.4K posts
- 6. #BravoCon 5,588 posts
- 7. RIP Coach Beam 1,265 posts
- 8. Marlene 5,138 posts
- 9. $EDEL 1,131 posts
- 10. Vatican 15.9K posts
- 11. Last Chance U 5,792 posts
- 12. Matt Taylor 3,054 posts
- 13. $GOOGL 20.2K posts
- 14. Wale 45.9K posts
- 15. Donica Lewinsky 2,606 posts
- 16. Oakland 11.6K posts
- 17. Crooks 79.3K posts
- 18. Samus 6,491 posts
- 19. Hunter Biden 23.2K posts
- 20. Freddie Mercury 1,363 posts
Something went wrong.
Something went wrong.