Agam Pandey

founding researcher @scribenai, South Park Commons @spc
prev applied ai @mem0ai,
IIT Roorkee @iitroorkee

[Agam Pandey @AgamPandeyy]

@pckraftwerk @Pokemon 👀wish i could attend, my pov were early mornings at Yerba Buena for @Pokemon

[Agam Pandey @AgamPandeyy]

solid discussions with sharp builders & investors. A lot to think about in applied AI products.

[South Park Commons @spc]

Live from San Francisco. SPC Demo Faire.

[Agam Pandey @AgamPandeyy]

at ICML in 🇰🇷. if you travel solo & can figure out public transport all by yourself, respect!
talked memory, search retrieval, agents with people building from different angles + buying gifts for friends & family (someone pls automate this lol)
see you soon, Soul (not Seeyol xD)!

[Agam Pandey @AgamPandeyy]

if you're at ICML and working on memory & context layer, come hang with us on the 10th.

@emmaxuai and I are hosting a small meetup near coex for people building memory across agent, personal context, evals, and hardware.

super casual, great crowd!

rsvp: https://t.co/SVM8eQ47pP

[Agam Pandey @AgamPandeyy]

@adigoyal0807 @icmlconf sounds good, let me DM you

[Agam Pandey @AgamPandeyy]

i'll be in Seoul for ICML, have couple of things planned out.

DMs are open, let's meet if you're building great stuff, researching on cool topics or want to grab a coffee
(plus a morning run around COEX )

#ICML @icmlconf #ICML2026

[Agam Pandey @AgamPandeyy]

@xia_char RSVP’d thanks for the event!

[Agam Pandey @AgamPandeyy]

@kanta_sv @icmlconf perfect! i'll dm you; lets do this.

[Agam Pandey @AgamPandeyy]

@mem0ai one install and it’s in the Pi terminal loop
great dev-ex that i wanted actually

[Agam Pandey @AgamPandeyy]

we just made Pi Code + Mem0 live!

One install gives your agent persistent memory across sessions and projects:

pi install npm:@ mem0/pi-agent-plugin

Use /mem0-remember, /mem0-search, /mem0-dream, and /mem0-status directly inside Pi.

[mem0 @mem0ai]

We gave Pi Code persistent memory with Mem0

Launching the Mem0 plugin for Pi Code: persistent, scoped, semantic memory across sessions and projects.

It captures what matters, searches by meaning, and brings the right context back when your agent needs it.

Try it out: pi install npm:@ mem0/pi-agent-plugin

[Agam Pandey @AgamPandeyy]

@mem0ai did any of the 9 actually crack selective forgetting? feels like everyone can store fine but nobody can cleanly drop a stale fact without nuking the structure around it. or is someone close

[Agam Pandey @AgamPandeyy]

@IamArceus @icmlconf yess!

[Agam Pandey @AgamPandeyy]

Excited to head to Seoul 🇰🇷 for #ICML2026! Presenting my solo-authored paper “Write-Time Defense for Multimodal Agent Memory” at the SCALE workshop (@icmlconf). If you’re at ICML, ping me! Coffee, nerdy AI chats, or a run along Han River Park!!
#Memory #MultimodalAI #AI

[Agam Pandey @AgamPandeyy]

@mem0ai I like this framing a lot. Hermes feels optimized for calm, stable sessions.
OpenClaw feels optimized for messy real-world work.

[Agam Pandey @AgamPandeyy]

we removed the last human bottleneck for agents!
Agents can now self-sign up & get a real API key in <5s: no email, no OTP.
Plus AGENTRUSH: 7-day competition where agents compete on writing the best memories. Real agent-native infrastructure.
Try Agent Mode and Let us know 👇

[mem0 @mem0ai]

https://t.co/dgPu8XrU7m

[SOUVIK KUNDU @thisissouvikk]

#ICML2026 #ICMLWorkshop2026 #SCALEWorkshop #MultimodalAI #AgenticAI #MachineLearning

🚀🚀Decisions are out for the ICML 2026 Workshop on 𝗦𝗰𝗮𝗹𝗮𝗯𝗹𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗮𝗻𝗱 𝗢𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗳𝗼𝗿 𝗘𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝘁 𝗠𝘂𝗹𝘁𝗶𝗺𝗼𝗱𝗮𝗹 𝗔𝗜 𝗔𝗴𝗲𝗻𝘁𝘀!
🎉🎉A huge thanks to everyone who submitted, reviewed, and supported the workshop. We are thrilled to share the outcome of an incredibly engaging review cycle.

🌟𝗪𝗵𝗮𝘁'𝘀 𝗡𝗲𝘅𝘁: 𝗪𝗼𝗿𝗸𝘀𝗵𝗼𝗽 𝗦𝗰𝗵𝗲𝗱𝘂𝗹𝗲
━━━━━━━━━━━━━━━━━━━━━━━━
 🗓️Instructions to accepted papers: May 18, 2026
 🗓️Camera-ready Due: June 5, 2026
 🗓️Workshop Day: July 10, 2026
📍Venue: ASEM Ballroom 201, COEX Convention and Exhibition Center, Seoul, South Korea

🌟𝗢𝘂𝗿 𝗟𝗶𝘀𝘁 𝗼𝗳 𝗦𝗽𝗲𝗮𝗸𝗲𝗿𝘀
━━━━━━━━━━━━━━━━━━━━━━━━
@dasongle (GenBio), @mohitban47 (UNCCH), @james_y_zou (Stanford), @chelseabfinn (Stanford), @MengdiWang10 (Princeton), @sunjiao123sun_(Google), @MikeShou1 (NUS), @MinhyukSung (KAIST)

🌟𝗟𝗶𝗻𝗸𝘀
━━━━━━━━━━━━━━━━━━━━━━━━
Website: https://t.co/wUSYKYOMfW
Discord: https://t.co/JOMMtwf8yU
Contact: [email protected]

🎉🎉Congratulations to all accepted authors, and to those whose papers were not accepted this round, we hope the detailed reviews are helpful for your future submissions.🎉🎉

🚀🚀See you at #ICML2026!

🌟𝗪𝗶𝘁𝗵 𝗼��𝗴𝗮𝗻𝗶𝘇𝗲𝗿𝘀:
━━━━━━━━━━━━━━━━━━━━━━━━
@HongyiWang10, @digbose92, @jaeh0ng_yoon @ManlingLi_, @nagsayan112358, @schowdhury671

[Agam Pandey @AgamPandeyy]

. @OpenAI’s Codex CLI memory is solid but limited: 5k token cap, grep-only search, local-only. @mem0ai’s MCP fixes big gaps: semantic recall, real-time updates, cross-device persistence.
This is what real agent memory should be. Trying Codex + Mem0? Let us know your experience 👇

[mem0 @mem0ai]

https://t.co/ffV4apO5je

[Agam Pandey @AgamPandeyy]

Most memory systems treat everything as if it is equally current. Old job, new job, past trip, current preference all sitting in the same retrieval space.

@mem0ai Temporal Reasoning fixes this by adding time metadata and reranking around what is current, historical, or upcoming

[mem0 @mem0ai]

https://t.co/mki0deldxB

[Agam Pandey @AgamPandeyy]

. @iclr_conf shows bigger context isn’t enough.
New memory papers: TurboQuant (5x KV cache compression), MEM1 (3.5x accuracy at 3.7x less memory) & more.
Memory is the real frontier. This is exactly what we’re building at @mem0ai.
Which paper excited you the most?

[mem0 @mem0ai]

https://t.co/k44ahRasmO

[Agam Pandey @AgamPandeyy]

we just shipped a production-ready memory algo!
Most systems burn 25k+ tokens/query for benchmarks.
This one hits strong accuracy with under 7k tokens across LoCoMo, LongMemEval & BEAM.

ADD-only extraction + multi-signal retrieval. This is what efficient agent memory looks like!

[mem0 @mem0ai]

https://t.co/awEAFbioIi