goofing around ✦ trying not to take things seriously ✦ ai product @razorpay
ahm/blr
- Tweets 1,306
- Following 964
- Followers 293
- Likes 2,598
can we have a catan for AI data centre?
you have the GPUs, i have the land, nobody has enough power, and one guy is hoarding water. honestly this may be too realistic to be a game.
@AnthropicAI @OpenAI @grok
somebody take notes for a merch
2
6
268
Riya Parikh retweeted
589497851845378089281879567916917641682963759302900576502946102 ± 1 are both prime
Happy weekend
65
111
4,141
358,955
Riya Parikh retweeted
It's becoming obvious that we are all spending a lot of time chatting with things that don't push back. Not good.
77
44
818
26,907
context is the moat. no wait evals are the moat. actually shipping is the moat. fuck it, a working sleep schedule is the moat. wait no. finding your airpods is the only moat.
speed is the only moat. no actually taste is the moat . wait no, urgency is the moat . fuck it, love is the moat. actually being normal is the moat. whatever. having a rent-controlled apartment is the only moat .
1
513
half the ‘smart agents’ i’ve tried are just slow to remember anything
Banger report on building faster memory for AI Agents.
If your agent's memory layer is slow, this design is worth a look.
Great weekend read.
It proposes a new agentic memory architecture, Jev-Mem, inspired by System-One/System-Two cognition.
This builds agent memory 6.6x faster and cuts query latency by 36.7% while scoring higher on LoCoMo.
In this system, a lightweight controller makes the memory decisions.
It assigns memory types and relations when storing, and handles routing, retrieval budget, graph traversal, scoring and stopping when reading. The LLM is called only for final reasoning and answer writing.
On LoCoMo, it scores 0.777 overall with an LLM judge, an 11.0% relative improvement over the strongest baseline. Memory construction takes 158 seconds, and average query latency drops to 0.93 seconds.
Paper: academy.dair.ai/papers/jev-m…
2
349
