Warp - The Agentic Development Terminal that Predates ChatGPT

I always know I’ve made a great choice to partner with a startup by how enthusiastic and invested the entire team is by the prospect of sourcing excellent early talent and that’s exactly how I knew I had hit the jackpot with Warp. Mac Bennett, Warp’s Talent Partner, and I met back in September where we instantly connected about the frictions of the talent industry and our mutual disapproval of LinkedIn, so when I proposed The Niche as an option for Warp to connect with some of the brightest talents, the team was immediately on board. As an avid user of Warp myself back in college, I was even more elated to have the opportunity to chat with Zach Lloyd, founder of Warp whose stacked career before the startup explains the team's high hiring bar and product-centric engineering culture.

The genesis for Warp came organically. As a Principal Engineer at Google working on Google Docs and Google Sheets, Zach had developed an appreciation for polished productivity software. Yet, the terminal, a developer tool he used practically every day paled in comparison to the user experience of other productivity software. "The experience was quite shitty," Lloyd recalls. "It puts a lot of onus on you to learn how to use it." Indeed, it wasn’t until a brief crash course in CS50 where I even learned how to use “ls” and “cd” in the terminal, and even after taking Systems, the terminal often felt foreign.

Predating ChatGPT

The biggest question in those early days though was whether you could build a business around something free that comes with every computer. The bet was that because the terminal was an incredibly valuable developer real-estate, there was a chance that people would pay by building a highly collaborative platform around it. The original pitch in 2020 pre-dated ChatGPT and was a platform where teams could share terminal sessions and work together through the command line, similar to how Postman transformed API development.

After nine months of dogfooding, the team launched a private beta on Hacker News in mid-2021 and received 10,000 signups on day one. They spent the next year and a half shaping the product experience on the platform: adding mouse support, structured output blocks, and reimagining the basic developer experience. It wasn’t until early 2022 after a year and a half of building the product that Warp fully launched, with an effort not to monetize up front at first.

What's incredible is that Warp predates ChatGPT. The company added collaboration features about 18 months in, and even experimented with simple AI (translating English to terminal commands) before ChatGPT existed. When ChatGPT launched at the end of 2022, Warp leaned harder into AI. The full product launched in early 2022, and the company has since pivoted to positioning itself as an "agentic development terminal.”

The Agentic Development Terminal

Warp is now a full-fledged AI-powered development environment that allows developers to interact with AI agents that can write code, execute commands, and navigate complex workflows, all within an interface that brings IDE-like capabilities into the terminal itself.

But with tools like Claude Code emerging, another AI coding agent embedded directly in the terminal, the competitive landscape is getting crowded. Moreover, Warp’s Agentic flows call frontier lab models like Claude. As a result, I asked Zach what Warp’s competitive advantage was here.

1. A Complete Terminal Environment, Not Just a Text-Based App

"When you run Claude Code, you're running a text-based terminal app," Lloyd explains. "With Warp, you're running an entire terminal." Warp can do everything Claude Code does, but it also offers richer user experiences that traditional text-based terminals can't such as traditional IDE features: you can edit code inline, conduct interactive code reviews where the agent sends code commenting directly within Warp, and the agent can seamlessly run tools like Vim or MySQL repls all within the same environment.

2. Model Agnosticism and Smart Pricing

Warp's second advantage is that they're not tied to a single model provider. "These models are constantly changing," and Warp supports bring-your-own-key (BYOK) functionality, giving users flexibility as the AI landscape evolves.

This model-agnostic approach influenced a fundamental shift in Warp's pricing strategy. Originally, subscriptions varied based on AI usage, a structure that hurt margins since their variable cost is inference. The new model combines a base subscription with usage-based credits. Warp also uses credit-based pricing rather than token-based pricing. "With tokens, we don't capture any benefit when we make things more efficient," Lloyd explains. “Credits, by contrast, are Warp's own currency.” One credit buys a certain amount of API usage, but not in 1:1 correspondence with tokens. If Warp makes their systems more efficient, they can capture that value rather than passing all savings directly to users.

Solo Founding

I was particularly interested in Zach’s journey as a solo founder. YC’s How To Found a Startup Podcast had drilled into me an aversion to solo-founding so I found Zach’s decision quite intriguing. Lloyd had co-founded a startup before Warp and understood the benefits of having a partner. But for Warp, there wasn't a specific person he wanted to build with, and more importantly, he didn't see how a co-founder would significantly benefit the venture at that stage. Sometimes the right founding structure wasn't about following conventional wisdom, but about honest assessment of what the specific opportunity requires.

A Technical Architecture Built in Rust

One of Warp's most powerful advantages is its visibility into the terminal, provided user consent of course. Because Warp sees what commands developers write, the company has access to incredibly rich behavioral data that could fuel a powerful flywheel effect. But Lloyd notes that data sharing is entirely opt-in. "It's up to the user if they're sharing data with us," he explains. "We have the potential for a full picture of what they're coding and doing in the terminal, but only if they consent."

For users who do opt in, Warp uses that data strategically across several areas: understanding engagement patterns (what do highly engaged users do differently?), improving conversion, and running AI evaluations to ensure their agents are succeeding or failing appropriately on tasks. Lloyd acknowledges they'd like to use the data for model fine-tuning as well, but they currently lack the engineering capacity. This is a vision for the company though especially as AI Coding tools like Claude break into the terminal.

Interestingly, Warp's entire codebase is written in Rust, an uncommon choice that reflects the team's performance-first philosophy. "We thought we could build a better product in Rust," Lloyd explains. "The terminal needs to be pretty high-performance and interface with the system very closely." The alternative would have been TypeScript/Next.js, which might have enabled faster initial development but would have sacrificed performance.

Rust's benefits for Warp are twofold: it builds faster apps by default, and it offers better support for cross-platform development. Most notably, the language choice became both a technical decision and a cultural signal, helping the company find engineers who were interested in learning and building in a new language.

The Team

The team composition skews experienced but deliberately includes new grads: roughly one-third have 1-5 years of experience, one-third have 5+ years, and one-third are recent graduates. The team is actively looking for new grads and/or interns where interviews often involve a chat with Mac, three technical interviews, followed by a product brainstorming session. For new grads, there is an additional technical and a bug bash. Additionally, we’ve linked the Notion page which deep-dives into their interview process!