Memory Store

everything everywhere, all at once

[Memory Store @memorydotstore]

@diwanksingh @IshitaJindal17 grug saved
memory store happy

[Memory Store @memorydotstore]

soo happy when the team loves and uses what they are building!!

[Ishita Jindal @IshitaJindal17]

i have severe dog allergy; and my claude knows what spots to avoid

we're out for a team offsite in Bali and were looking for a good coffee spot today morning.

i started a claude chat and asked for coffee shop recommendations. claude called memory store and it ruled out spots that have pets.

[Memory Store @memorydotstore]

@sridharfyi @ycombinator agreed @sridharfyi

[Memory Store @memorydotstore]

@HamadaSalhab @diwanksingh @pioneer_fund @jamesfong @CryptpalSingh done @HamadaSalhab!!

[Diwank Singh Tomer @diwanksingh]

@memorydotstore x @pioneer_fund cookies!

a befitting start to a sweet partnership! 🍪🤓

thanks @jamesfong @CryptpalSingh and team! 🫶

[Memory Store @memorydotstore]

@jamesfong Really excited to have @pioneer_fund as partners!

[Memory Store @memorydotstore]

4th best YC startup this batch?

[Fluenta @FluentaSpace]

YC Demo Day, June 16

The biggest stage in startups opens in four days. The Spring 2026 batch puts 194 companies on stage - https://t.co/WQY1EvYyTC. We scored all 194 before they walk on, then split the batch into four groups and we're dropping one a day.

The method: 6 public signals, a 0 to 100 Launch Readiness Score (LRS). Search demand, social pain, competition, monetization, funding, urgency. Public data only. The outside view every investor in that room runs before the meeting.

Group 1 of 4: Agent Infrastructure. The picks and shovels. 55 startups building the rails everyone else builds agents on. Group average: 50.3.

Highest scored:
1. Kuli, 65.5, automated influencer marketing - @maradoh22
2. Armature, 63.2, product analytics for agents - @Totzenberger
3. Superlog, 63.2, self-healing logging - @superlogYC , @ArseniySvist , @nicolomagnante
4. Memory Store, 62.4, memory layer for agents - @memorydotstore , @IshitaJindal17 , @diwanksingh
5. Netter, 62.4, data ops for mid-market

Lowest on the outside view:
51. Hub xyz, 38.2, real-world AI datasets - @hubxyz, @xarmin @tim404x
52. The Company Company, 38.2, autonomous company - @thecompanyai, @juliuslip
53. General Instinct, 38.0, physical-AI deployment - @BillJiao930 @Guanming717
54. Minicor, 37.7, self-healing desktop agents - @minicor_ , @faizchishtie @sahee_d
55. RentAHuman, 32.7, real-world tasks marketplace @rentahumanx, @AlexanderTw33ts

If the score reads low, it usually means the data on 6 public signals (search demand, social pain, competition, monetization, funding, urgency) is either poor or hardly available. Could be worth closing that gap on stage on June 16.

3 things the data says about this group:
1) Some are building ahead of demand, some aren't, and search alone won't tell you which. 33 of 55 sit under 500 searches a month (nobody googles "agent memory" yet). But a few have real pull: Kuli and Netter clear 11k at 80 to 94% purchase intent, and the biggest raw number, Runtime's 311k, is mostly generic "runtime" spill, not category demand. The caveat that matters most: this is public search only. Infrastructure sells through private B2B motion, so a low score can sit on top of signed LOIs, paid design partners, or institutional pre-orders the outside view can't see. If that's you, that private demand is the single strongest thing to put on stage June 16.

2) One bet, 55 ways, and from the outside every idea could have the same shape. Half literally build "for agents," 71% sit in dev tools or AI automation, median 10 direct competitors each. Strong willingness to pay, crowded competition, and weak funding, not because these founders can't raise, but because the capital already came to some of these categories in 2020 to 2022 and got absorbed, not broken out. The batch converged on one thesis and now competes with itself.

3) The strongest signal is the one that lies. Monetization scores highest across the board, people pay. But that score measures whether a market pays, not whether you can afford to win it. On the real CAC vs payback math, 14 of 49 priced startups need 12+ months to earn back one customer, and 13 of those 14 scored STRONG on monetization. So the question isn't "is there a business," it's "is this a company or a feature." That's what you get pushed on June 16.

How to use it:
- Founders in the batch: find your card, look at your lowest of the 6 signals. Those could be the questions coming. Worth prepping now.

- VCs and angels: want the full breakdown on any project, competitors, the CAC vs payback math, search demand, etc? Check on https://t.co/lM4X8lkhCH. DM us to uncover the paywalled parts as well. The reports are done, so no additional costs here. Or use Fluenta MCP to pull all reports in one batch and do a thorough analysis on each one in Claude, Cursor, etc.

- Everyone not in the room June 16: the high-LRS names are the ones who’d mostly likely get funded, and who get cloned in every local market inside 90 days. Browse the board for your business inspiration.

Tomorrow, Batch 2 of 4: The AI Workforce. The agents coming for entire job functions.

Tagging the teams above in case you want to grab your report. No hidden subscription, no signup gate - the analysis is already done, the file is yours to pick up. DM here or email [email protected] and we'll send it same-day.
If the score reads you wrong, that gap is the thing to close on stage.

[Memory Store @memorydotstore]

@superset_sh @zjearbear @ArgaLabs @totalistrading @heyclicky @archer_money @manicule @sazabi i 🩷 you too!

[Jerry Zhang @zjearbear]

YC P26 startups I’m bullish on:

>@ArgaLabs
>@superset_sh
>@totalistrading
>@heyclicky
>@archer_money
>@manicule
>@sazabi
>@memorydotstore
> Prism (bro better get on x)

Good luck at demo day 🤝

[Diwank Singh Tomer @diwanksingh]

Almost every memory in @memorydotstore has a timestamp. if the model formats the date wrong, the memory is stored incorrectly - and it’s super hard to catch.

So we benchmarked which datetime format LLMs actually get right: 22 models, 235 scenarios, 7 formats.

tl;dr:
- use RFC 3339. python datetime also works well.
- do NOT use JS Date or unix epoch directly - JS Date got parsing wrong 1 in 4 times,
epoch dropped to ~40% on time arithmetic.
- need epoch? have the model output a string timestamp, convert in code.

also: Sonnet beat Opus.

Opus overthinks it.🤷

[Diwank Singh Tomer @diwanksingh]

excited for @memorydotstore to be featured on @Forbes by @dasha_shunina! 🥰

[Memory Store @memorydotstore]

I made it to Forbes :)

[Dasha @dasha_shunina]

Meet the most promising companies from the latest @ycombinator batch 👀

One big takeaway:

AI agents need more than intelligence.
They need memory, identity, observability, compliance, insurance and power.

My latest @Forbes deep dive 👇

https://t.co/1IPkx2hqSi

[Ishita Jindal @IshitaJindal17]

Our @ycombinator launch video went viral.

- 90 demo calls booked
- 3x revenue in the last 2 weeks
- 150+ new people added to the waitlist

We only started monetizing 3 weeks ago, but we're getting a clear signals.

Teams don't want another personal memory tool, they want one shared brain their whole team and their agents can pull from.

[Ishita Jindal @IshitaJindal17]

@eladgil been talking to companies about this.

there are still lot of bottlenecks. and the biggest one is the way decision making works in orgs.

the differential between companies realizing this and adopting v/s the ones that are not, is huge.

[Divyanshu Rajput @DivyanshuRaj81]

Well, now life seems to be easy, thanks to @memorydotstore

[Madhavan (Maddy) @madhavanmalolan]

Previously i had a dump of all my slack messages and gave my agents a grep tool.

For a long time the grep tool got shit done.

But when memory store started referencing old conversations with abstract search terms - it gave me superpowers.

[Madhavan (Maddy) @madhavanmalolan]

The reason this is so useful is because, most ideas at our company have been discussed before in some form or shape.

It is hard to refer to them, or even remember that we had talked about this. But @memorydotstore consitently digs out gold.

[Madhavan (Maddy) @madhavanmalolan]

. @memorydotstore might be the latest ai tool that crossed the novelty chasm into actually being useful.

Most ai tools i used off late are hard to use beyond the first couple demo runs.

The others in this category are - codex, claude design, (far third) openclaw. S tier.

[Memory Store @memorydotstore]

i love making users feel like a superhero,

[Ishita Jindal @IshitaJindal17]

you don't notice a memory tool filling up

you notice the day it's gone

@madhavanmalolan didn't see the value at first - i actually took him off my lead list.

weeks later he sent me this 👇

[Memory Store @memorydotstore]

@a3fckx @opencode @grok @CommandCodeAI @NousResearch you should do video tutorials on how to set up agents :D