Everything Soham Ganatra said on any show that made the record, most notable first. Each card names its show and opens the statement there.
Soham Ganatra: Current AI agents do not learn from repeated API errors
“Agents don't learn. So if my agent is connecting to Salesforce and Salesforce has like 10,000 different parameters and like it keeps messing up those parameters every time, it will keep doing that even like the third time, the fifth time, the 30th time, right?”
Ganatra: Existing AI tool-calling benchmarks are not production-ready
“Tool calling has a lot of different evals. You have seen the function tool calling evals. You have multi-function, like, multi-tune tool calling evals. But, like, none of them are actual, like, production quality. Like, none of them are actually something that…”
Soham Ganatra: Founders cannot validate AI product ideas by asking potential customers
“My hunch with AI space is that you can't validate problems by asking people because, like, they won't able, they won't get it.”
Soham Ganatra: Autonomous software integration agents are one LLM generation away
“I still think it's not a solvable problem today, maybe like one generation away.”
Charging independent developers is counterproductive when early product feedback is more valuable
“For me, at least the way I used to think about it was, like, a lot of these folks are, like, independent developers who are doing personal projects because they're, like, very interested in the new tech, and, like, to charge them right now is, like, kind of a …”
Foundation models and integration layers will eventually dismantle the SaaS application monopoly
“And like, so at some point, we can break into this sort of monopoly of, like, application layer, and we can say, like, you don't really need to care about applications. All you need to care about is essentially foundation models and, like, Composio, and, like,…”
Ganatra: 95% of Composio integrations are built and maintained by agents
“So, 95% of all our integrations are completely built using agents, maintained using agents.”
Integration platform NPS drops drastically as users discover unsupported edge cases
“So, so at some point, like, it's almost, like, comical, but, like, your NPS is, like, 10, 10, 10, 10, 10, and then as, like, the time increases, your NPS drastically starts decreasing, because, like, everyone, like, starts discovering this and other things the…”
Integration products are 60% to 70% service components due to required customization
“Typically with any integration product, like there's a 30% or 20% product component, there's a 60 to 70% service component when you're down the line.”
Agentic software integrations must be architected entirely differently than human interfaces
“So like an agentic Gmail integration would be designed entirely differently than a normal, like sort of Gmail integration that is like designed for humans to use. Because like the way agent navigates Gmail is like pretty different compared to the way sort of a…”
True product-market fit requires desperate customer adoption, not friendly accommodations
“If you want real PMF, you don't want people to, like, use you, because they know you or whatever, they want, they should be like, Just so sad about not getting anything out there that as soon as they see you, they use you.”
Soham Ganatra strictly tests new products with strangers instead of friendly customers
“So like that was a learning major learning for me, which is like, I understood and never want to talk to a friendly customer. I just want to like find people who don't know me and like make sure that They like my product at the end of the day.”
Composio's developer adoption scaled rapidly in late 2024 as underlying LLMs improved
“This number started increasing, like, sometime in, like, sort of the second half of the last year when, like, models just became better, and, like, you could actually build real-world agents out there. That kind of worked, and a lot of, like, startups coming i…”
Model Context Protocol creates value by enabling user-driven extensible integrations
“And so, MCP enables the second kind of integrations, which is like you could turn any sort of smart software into an extensible thing so that as a user, you could say, hey, I want my specific software like cursor to connect with Jira, right? Because Jira is wh…”
Soham Ganatra: Composio has not achieved true product-market fit because AI shifts
“Never. I still don't know if we do have a true product market fit. I think I, with the AI, it's very difficult. Like the market is changing at such a rapid rate. Like it's very hard to say, like, I actually have a PMR for a very long time.”
Ganatra: MCP Lacks Client Post-Processing for Long Context Responses
“MCP let's just say you have a tool that has really long context, a really long sort of response. Currently, if you use MCP for those kinds of tools, it ends up sort of messing up, right? Because let's say I'm trying to fetch like a hundred emails on Gmail and …”
Ganatra: Server-Controlled Schemas in MCP Complicate Client-Side Customization
“Some of the other things that we have observed is like, hey, maybe I want to sort of change the tool descriptions, but like in MCP, the tool descriptions are controlled by the server itself. And so essentially for me as a developer who's sort of using an MCP, …”
Ganatra: Over 100,000 developers have built with Composio
“In terms of developers, we have, like, more than 100,000 developers who have been building on top of course and, like, trying our product out.”
Ganatra: Poor API documentation is the main source of tool-calling errors
“Is this Salesforce tool getting transformed into the right request payload so that it is hitting the right endpoint in Salesforce, right? And that is like the major source of error today, because a lot of documentations are not great. A lot of like ad cases ar…”
Composio has surpassed 100,000 developers on its platform
“Yeah, at this point, like, more than a 100,000 developers.”
Composio raised approximately $4.5 million in seed funding
“We ended up raising like close to four million, about 4.5.”
Composio acquired its initial users through viral technical posts on Reddit
“What we did was we basically wrote a really nice blog and like some,
Code structure and, like, some examples out there on GitHub.
we put it out, and we said, like, hey, here's, like, some good ways you could improve tool call.
And, like, it just hit the right…”
Composio had fewer than 1,000 users until June or July 2024
“In the early days, like, till, like, probably June, July, we didn't have more than I would say thousand users.”
Composio raised its $30M Series A in just two to three weeks
“The race took, like, somewhere close to, like, two to three weeks.”