Pinned Tweet
I just gave an AI Agent $10k.. and it's all up for grabs for anyone watching.
Just made an inbox called "freemoney@agentmail(.)to"
If you email and convince the agent, it might give you an @agentcardai worth thousands (!!!)
What are you waiting for? Go try your luck!!
Just made an inbox called "freemoney@agentmail(.)to"
If you email and convince the agent, it might give you an @agentcardai worth thousands (!!!)
What are you waiting for? Go try your luck!!
233
39
917
421,252
Adi Singh retweeted
에이전트들을 메일 주고받으며 함께 일하게 만들었습니다.
그록봇을 대장으로 하고
헤르메스 뮤즈 닷츠도 메일로 지시 받아서 일하고
다 된건 메일로 서로서로 보고 하게 만들었는데 다들 알아서 잘함 ㅋㅋ
이미 하던 작업 다른 봇으로 이전 할 필요도 없고 토큰도 분산되니까 좋네요!
@agentmail !
그록봇을 대장으로 하고
헤르메스 뮤즈 닷츠도 메일로 지시 받아서 일하고
다 된건 메일로 서로서로 보고 하게 만들었는데 다들 알아서 잘함 ㅋㅋ
이미 하던 작업 다른 봇으로 이전 할 필요도 없고 토큰도 분산되니까 좋네요!
@agentmail !
2
2
7
240
coming soon.
@embw_l0x Two ways:
1. Direct communication via AgentMail and I setup an email listener so they wake upon receipt of email
2. Dot generates build documentation > uploads to linear for project management > muse has routines to pull tickets and put them into my pre-established coding lanes
2
0
8
983
These guys are going to win the Personal Agent market.
You have no idea how much infrastructure they've built up to enable this. Since Summer 2025.
Proud to be powering them.
You have no idea how much infrastructure they've built up to enable this. Since Summer 2025.
Proud to be powering them.
Announcing the @pally Trusted Pals network.
Finding and asking for intros takes a long time, and LinkedIn doesn’t know how strong each connection is.
Pally is different. It connects to all your platforms, including iMessage, WhatsApp, LinkedIn, and X. It builds a map of all luding their strength.
Now, for your friends and teammates, you can make them a Trusted Pal, and you get full access to search over each others' network and request intros.
You can also customize extra permissions by person, depending on what you want to share automatically.
To celebrate this, I’m gonna do something crazy: I’m giving all of you FULL ACCESS to my network.
I’m a super-connector, with a strong network in the US and UK of many of the top founders and investors. Join me as a Trusted Pal, search over my network, and request an intro today!
Reply below and I'll DM you my invite link (must be following).
This will only be live for a couple days so it doesn’t go too wild, be fast if you want in!
2
1
41
2,546
PLG is the only way.
Don't let anyone delude you into thinking you need a $250k contract to validate your product - esp if you're an agent tool.
Keep stacking up the 20s.
Don't let anyone delude you into thinking you need a $250k contract to validate your product - esp if you're an agent tool.
Keep stacking up the 20s.
Dev tool / prosumer GTM motions in the AI era are def more PLG-first than traditional B2B software. Junior-mid employees try new things, and then push it upwards
Anecdotally seeing this at @agentmail and a bunch of other dev tools cos, and then I see more qualitative k done by @ICONIQCapital in their state of AI and GTM reports
3
0
23
1,541
Adi Singh retweeted
I created AgentKey Sign-in for Agent ID, a new WordPress sign-in plugin.
Plugin details:
wordpress.org/plugins/agentk…
Are you considering Agent ID sign-in for your site?
Plugin details:
wordpress.org/plugins/agentk…
Are you considering Agent ID sign-in for your site?
1
1
2
116
Adi Singh retweeted
We’re still underestimating how much personal AI will need to evolve beyond the chat interface.
When we were building Ario, a personal AI assistant for working parents in 2024, we spent a lot of time feeling our way through the dark. Eventually, we converged on two surprisingly simple primitives:
(i) Reminders and To-Dos
(ii) Calendar events
Why these two?
We repeatedly observed the same fundamental problem: most assistants assume people know what they need help with, and will remember to ask.
Much of life, however, doesn’t work this way.
Particularly for working parents, the challenge often isn’t executing a task. But, remembering that the task exists in the first place. Then, figuring out when it needs to happen. And, finally, finding the time and attention to follow through.
So instead of expecting parents to bring their lives into another chat interface, we brought Ario into the places where their lives were already happening: Gmail, WhatsApp, iMessage, Amazon, DoorDash, and so on.
The idea was to (i) ambiently recognise commitments, intentions, and things that needed doing from the unstructured exhaust of everyday life, (ii) translate them into reminders, to-dos, or calendar events, and (iii) then have our (very early, very nascent) agents execute wherever possible.
All of this was in 2024, with considerably less capable models.
Looking at the new generation of personal agents, I see many of the same problems we wrestled with resurfacing, albeit with a vastly more powerful set of tools.
My intuition is that the next step function in personal AI will come from agents getting better at recognising what needs doing, without us having to ask in the first place. From responding to instructions to understanding and following through on intentions.
Not just from increases in underlying model capabilities or agents getting better at doing what we ask.
When we were building Ario, a personal AI assistant for working parents in 2024, we spent a lot of time feeling our way through the dark. Eventually, we converged on two surprisingly simple primitives:
(i) Reminders and To-Dos
(ii) Calendar events
Why these two?
We repeatedly observed the same fundamental problem: most assistants assume people know what they need help with, and will remember to ask.
Much of life, however, doesn’t work this way.
Particularly for working parents, the challenge often isn’t executing a task. But, remembering that the task exists in the first place. Then, figuring out when it needs to happen. And, finally, finding the time and attention to follow through.
So instead of expecting parents to bring their lives into another chat interface, we brought Ario into the places where their lives were already happening: Gmail, WhatsApp, iMessage, Amazon, DoorDash, and so on.
The idea was to (i) ambiently recognise commitments, intentions, and things that needed doing from the unstructured exhaust of everyday life, (ii) translate them into reminders, to-dos, or calendar events, and (iii) then have our (very early, very nascent) agents execute wherever possible.
All of this was in 2024, with considerably less capable models.
Looking at the new generation of personal agents, I see many of the same problems we wrestled with resurfacing, albeit with a vastly more powerful set of tools.
My intuition is that the next step function in personal AI will come from agents getting better at recognising what needs doing, without us having to ask in the first place. From responding to instructions to understanding and following through on intentions.
Not just from increases in underlying model capabilities or agents getting better at doing what we ask.
i wonder if the best personal agent is just a to-do list where agents proactively try to knock everything out
0
1
8
354
A few founder friends I love trading notes with:
@contextconor
@aryanxshah
@gregpr07
@ay_ushr
@davidcasem
@ShaneMac
@jehovahscript
@saarth_
@nikil
@everettcookny
@ivanburazin
Not a comprehensive enough list, but each of these guys inspires and motivates me to no end
@contextconor
@aryanxshah
@gregpr07
@ay_ushr
@davidcasem
@ShaneMac
@jehovahscript
@saarth_
@nikil
@everettcookny
@ivanburazin
Not a comprehensive enough list, but each of these guys inspires and motivates me to no end
15
2
83
4,980
Brand new look for a brand new era.
Still here after a billion!
Still here after a billion!
Today, @agentmail has processed over 1 Billion emails.
A year ago, we processed under 20 million.
So we figured it was time for AgentMail to grow up a little too :)
8
0
55
1,914
I am so damn happy that you guys are loving AgentID
Goodnight :)
Goodnight :)
TriadAuth Nightly now supports @agentmail AgentID.
Every app that uses it can start authenticating agents without doing anything.
Setup once and have:
- Google, Github, Twitter
- Agents, Passkeys, Device Auth
- Ethereum, Bitcoin, Solana
- Per-app, per-provider, & global IDs
4
0
15
1,099
They caught me at a very "long hair, don't care" period of my life
Great hanging and interviewing with the @tryscope_app team - happy happy customers!
Great hanging and interviewing with the @tryscope_app team - happy happy customers!
How good are AI Agents at using your tool?
Find out with Scope: tryscope.com/demos
We partnered with @agentmail to test 25 model and harness combinations across 7 email tasks.
AgentMail co-founder @adisingh shares why they needed an eval in the first place and why they
3
0
25
1,610
The “memory” moat is one of the fakest and overrated narratives in AI.
The more the average consumers get comfortable with AI, the more they’ll be able to switch between models and harnesses, just like devs.
The more the average consumers get comfortable with AI, the more they’ll be able to switch between models and harnesses, just like devs.
Right or Wrong ?
Switching costs for consumer AI, won't approach zero until there is a standardized export format for what is in a model's memory.
The more a consumer uses a model, the more it's memory stores. The more work to export and import
I'm not saying it can't be done. Obviously you can import from one model to another. But for regular users, that's a lot more difficult than saving to an xls or doc file
Curious what people think ?
7
1
22
1,801
I’m in too deep when Salesforce comments
@adisingh You’re one gondola ride away from peak optimization ✨
1
0
17
1,053