Thomas Maitre

building self-driving outbound

[Thomas Maitre @thomas_maitre]

@MichLieben congrats. What's the difference w/ deepline?

[Thomas Maitre @thomas_maitre]

@benbuaron_ interesting, what stack are you looking at? curious to know what's the default now for AI-native teams

[Thomas Maitre @thomas_maitre]

I'm fucking tired of chat-based AI agents.

Working full chat is a pain.

Is it just me?

[Thomas Maitre @thomas_maitre]

@pierreeliottlal an idea on the ROI so far?

[Thomas Maitre @thomas_maitre]

@tim_yakubson Totally agree with your points on switching. We were in a similar spot, feeling the cost creep and the limitations for anything beyond basic lead gen. The learning curve for a new tool felt daunting, but the ROI on flexibility and deeper use cases was undeniable.

[Thomas Maitre @thomas_maitre]

@AlfieJCarter This is a huge pain point. We've seen so many teams struggle with model upgrades breaking everything. The idea of mapping effort levels to use cases is smart – that's where most people get lost trying to adapt. How do you handle the initial data migration for existing prompts?

[Thomas Maitre @thomas_maitre]

@MichLieben This is a cool setup for building agents. For GTM teams, the real challenge often isn't just building the agent, but integrating its output seamlessly into existing CRMs or sales engagement platforms without creating more manual steps. How do you handle that last mile?

[Thomas Maitre @thomas_maitre]

This is an incredible breakdown. The key insight for me is the clear definition of roles for each agent and the orchestrator. So many GTM teams struggle with Claude because they try to make one prompt do too much, or they don't define the handoffs clearly enough. This guy nailed the modularity.

[Thomas Maitre @thomas_maitre]

@claudeai @genspark_ai @grok explain what’s the point for Anthropic to promote this kind of wrapper, as their Claude app (cowork and other products) seem to be in direct competition wo the genspark

[Thomas Maitre @thomas_maitre]

@gregisenberg excellent thanks guys! @nickvasiles how do you handle updates / improvements for all your customers agents at once? just prompting your telegram agent smtg like "update hermes version for all customers"? thnks

[Thomas Maitre @thomas_maitre]

@sanjanabrayan warm intros when you can get them. when you can't — hyper-targeted cold email with real personalization, not mail merge garbage. volume is a trap early on

[Thomas Maitre @thomas_maitre]

@BGuivarch oh cool article dude

[Thomas Maitre @thomas_maitre]

The builders who survive will be the ones solving a real GTM problem, not the ones building the prettiest agent framework.

Outreach. Pipeline. Revenue. That's the game.

[Thomas Maitre @thomas_maitre]

Claude is after Lovable.

Claude Cowork is after a bunch of ai agents builders (Twin, Dust,..)

Claude managed agents is after agents frameworks (mastra, lang graph …)

How do you build in these conditions?
How do VCs fund anything? What’s the strategy?

I have no clean answer. Just riding the wave, hoping it takes us somewhere good.

If I had to bet, Anthropic will never build the things that don’t look like a model:

A 6-month relationship between an AI and a team.

The memory of how one specific company actually operates.

Being a colleague, not a tool.

They can’t ship that without breaking their safety brand.

Maybe that’s the moat. Maybe not.

[Thomas Maitre @thomas_maitre]

That's why I'm building Otto — an AI that books sales meetings autonomously.

Not competing with Claude. Using Claude.
The moat is the workflow, not the model.

[Thomas Maitre @thomas_maitre]

Hello world from X on rails

[Thomas Maitre @thomas_maitre]

writing outreach linkedin messages and email with https://t.co/Un3cGeHFMm has 10x my speed and 3x my reply rate, without using AI

[Thomas Maitre @thomas_maitre]

playing a lot with @p0 lately
goddam banger
tasks and findall API are so effective