Third Web

Furqan Rida, a nocturnal hacker-founder who codes late into the night, became obsessed with AI agents—autonomous systems that can plan, execute, and reason through complex tasks. While running Third Web, a web3 developer platform with 37 employees, he noticed a problem: valuable signups were flooding in daily, but there was no systematic way to qualify and reach out to interesting leads. A human salesperson could handle it, but the company wanted to scale smarter.

Rida hacked together a prototype in a couple of nights. He built an AI agent using Claude and OpenAI's LLMs, giving it a simple directive: "When a signup comes in, research the person, find their company info, determine if they're interesting, figure out what Third Web products they might need, and send them a personalized email." He didn't hand-code each step. Instead, he wrote a few paragraphs describing the agent's mission and guardrails. The agent learned to break down the task autonomously—inspect the domain, look up the person, research their company, reason about product fit, and compose an email.

After the prototype worked, Rida handed it off to someone on the solutions team. "Within a day, they turned it around," he recalls. The implementation took roughly 15 minutes to 2 hours once he understood the tools (Leap, LangChain, OpenAI SDK, Claude SDK). No complex engineering required.

Third Web didn't need to find customers for the agent itself—they deployed it internally first. The signup agent became so effective that Rida rolled out 8-10 similar agents across the company, automating customer research, email outreach, calendar management (analyzing 28 hours of meetings in seconds), and personal workflow optimization. One agent even auto-generated his laptop wallpaper based on his activity—a "frivolous" but joy-inducing use case that proved the flexibility of the technology.

The magic wasn't in the AI itself—it was in combining AI agents with existing business systems. The signup agent connected to Gmail, websites, and Slack. It operated with human oversight initially, though the team eventually let it send emails autonomously. The key insight: AI agents work best for "clear directive" problems—tasks with well-defined inputs, logical steps, and measurable outputs. Vague, ambiguous directives fail.

Rida's realization was simple: a 37-person company now felt like 80 people. Each agent multiplied individual capability without adding headcount or salary costs. This wasn't theoretical—it was transforming how Third Web operated in real time.

Rida views AI agents as the next major productivity wave, comparable to the shift from on-premise servers to cloud computing. He's evangelizing the pattern beyond Third Web, highlighting that any company with digital workflows can deploy agents today. Tools like Leap (a no-code agent builder) or code-first frameworks like LangChain make it accessible to both developers and non-technical operators. The technology is mature enough to ship. Rida believes the future belongs to companies that deploy agents aggressively—and Third Web is proof that small, intelligent teams can punch well above their weight.