Jun 25, 2025 · 52m · latent-space
⚡️Warp 2.0: the Agentic Development Environment - Zach Lloyd and Ben Holmes
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Warp CEO Zach Lloyd and Ben Holmes join the Latent Space podcast to launch Warp 2.0, showcasing its transformation from an AI-native terminal into an Agentic Development Environment with multi-threaded autonomous coding, rich developer UX, and enterprise-grade workflows.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 17% of the talking time here. How this is scored →
The hosts as informed peer 5.3 Guest teaching 3.7 Guest disagreement 1.6 The hosts pushing back 2.7
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 35:23 Zach counters the existential threat of AI labs
Zach directly rejects the notion that foundation labs have an insurmountable lock-in, emphasizing that model-agnostic layers like Warp can switch to whichever lab takes the performance lead.
Hardest push from the hosts ▶ 25:56 Alessio pushes back against terminal lock-in for Warp CodeAlessio directly challenges Warp's bundled architecture, asking why developers who prefer other terminals like Ghostty should be forced into the Warp terminal UI just to use Warp Code.
Biggest teaching moment ▶ 29:39 Zach explains developer rejection of remote cloud sandboxesZach walks through years of failed attempts to push developers to remote sandboxes like Codespaces, detailing the practical setup friction and hardware constraints that keep development local.
The host holds their own ▶ 44:48 Swyx challenges the engineering overhead of production GraphQLSwyx presses the Warp team on whether their production GraphQL backend actually justifies the abstraction overhead compared to modern type-safe RPC frameworks.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Introducing Warp 2.0 and Guests Zach Lloyd and Ben Holmes | 4 | 2 | 1 | 1 | Swyx introduces Warp from personal experience switching from iTerm2. Zach introduces Warp 2.0's positioning as an agentic development environment and benchmark goals. | |
| From Human-in-the-Loop Fixes to Autonomous Agent Profiles | 5 | 3 | 1 | 2 | Swyx recounts his earlier critiques and experience hitting enter repeatedly to fix Python dependencies in Warp. Zach validates this and explains how agent profiles now allow autonomous execution. | |
| Why Warp Built an Integrated, End-to-End Agent | 6 | 4 | 2 | 3 | Alessio asks why Warp built an end-to-end coding agent instead of delegating to ChatGPT desktop. Zach explains the need for deeper read-write CLI integrations, prompt control, and multi-model flexibility. | |
| Demo: Multi-Threaded Agents and Sentry Bug Resolution | 3 | 4 | 1 | 1 | Zach demonstrates speech-to-text prompt input, linear issue context retrieval, and multi-threading a second agent to debug a Sentry Go null pointer bug. Hosts observe the high-effort prompting workflow. | |
| Demo: Reusable Saved Prompts, Inline Diff Viewer, and Code Editor | 4 | 3 | 1 | 1 | Zach demonstrates saved team prompts for Cloud Run log analysis, inline diff viewers, and the newly built code editor inside Warp. | |
| Demo: Production Log Analysis and Cloud-Synced Rules | 3 | 3 | 1 | 1 | Zach shows live staging log analysis and explains Warp's cloud-synced rules engine compared to local Cursor rules. | |
| Demo: Natural Language in Postgres REPL and Wrap-Up | 4 | 3 | 1 | 1 | Zach demos natural language SQL querying inside a Postgres REPL session and wraps up the live product demonstration. | |
| Rethinking Developer UX: Terminal Abstractions vs. Chat Panels | 6 | 4 | 2 | 3 | Alessio asks about legacy terminal baggage versus building an agent-first UI. Zach and Ben discuss why terminal abstractions like tabs and panes map cleanly to agent containers compared to chat panels. | |
| The Case for Native Agent UX over Headless CLI Tools | 7 | 5 | 3 | 6 | Alessio challenges Warp's bundled approach, asking why users should not just use Warp Code in Ghostty or headless CLI. Zach counters that native UI for editable diffs and in-tool code reviews provides essential UX value. | |
| Local Workflows vs. Cloud Sandboxing in Multi-Agent Development | 7 | 6 | 2 | 5 | Swyx presses on agent sandboxing, pointing out terminal commands produce side effects that interfere across multi-agent setups. Zach acknowledges cloud sandboxing is necessary long-term but defends starting locally for fast feedback loops. | |
| Defining the Agentic Workbench and Competing with AI Labs | 7 | 5 | 3 | 6 | Swyx challenges Zach on competing directly against foundation model labs with massive resources. Zach articulates Warp's positioning as an agentic workbench with model agility. | |
| Warp Growth Metrics, Market Shifts, and Enterprise Adoption | 5 | 3 | 1 | 2 | Swyx asks for growth metrics. Zach shares that Warp has 600k active users and 5-15% weekly revenue growth, noting a market shift toward enterprise willingness to pay for developer AI. | |
| The Economics of AI Developer Tools: Subscriptions vs. Usage Pricing | 6 | 4 | 2 | 3 | Swyx and Alessio discuss the tension between flat-rate subscriptions and usage-based API pricing in dev tools. Zach details Warp's tiered plans and the psychological barrier of per-call meter pricing. | |
| Enterprise Collaboration, Knowledge Bases, and Security Gateways | 6 | 3 | 1 | 2 | Swyx asks about enterprise security and gateway controls. Zach explains how Warp's shared knowledge base, MCP configurations, and enterprise LLM proxy routing integrate with enterprise security requirements. | |
| Architectural Insights: Production GraphQL and Go Backend | 7 | 4 | 3 | 5 | Swyx and Alessio express surprise at seeing production GraphQL logs, questioning its value over simpler RPC/REST. Ben and Zach defend its utility for dynamic tool call schemas while acknowledging backend Go struct friction. | |
| Warp's Founding Story and Dylan Field's Investment | 5 | 3 | 1 | 1 | Alessio and Swyx ask about early investor backing by Dylan Field. Zach shares the founding thesis and background as former principal engineer on Google Docs before closing the episode. |
Statements from this episode (30)
Zach Lloyd: Warp 2.0 launches as an agentic development environment
“We are today, I think as this podcast comes out, releasing warp two, which, you know, we're positioning ourselves as being no longer just a terminal, but an agentic development environment.”
Zach Lloyd: Warp's coding agent will likely top the TBench benchmark
“Basically, state of the art on SweetBench, I think we will, again, I don't want to be quoted here, we can maybe edit this later, but like, we'll probably be number one or close to it on TBench also, which is the terminal benchmark, which really we should be th…”
Zach Lloyd: Warp introduced Agent Profiles for granular autonomy control
“We've introduced the ability to let that go fully agentically now. So You know, there's, like, a new set of, we call it, like, an agent profile, where you can, at the level of, like, running commands, or executing diffs, or doing code search, you can control t…”
Lloyd: Warp builds native agent because tool-calling integration goes deeper
“The main reason is like, we feel like the integration can go way, way deeper. If warp itself is the thing driving the interaction primarily because of like the, Type of tool calling that we can do, like, directly in Warp.”
Zach Lloyd: Anthropic, not OpenAI, is Warp's primary model of choice
“ChatDBT isn't even the model of choice at this point. It's anthropic for us.”
Lloyd: The best agent interface is similar to traditional terminal UX
“Our thesis here is that the best interface for launching agents is actually pretty similar to what traditionally has been like a terminal UX. You know, you want this like grouped output. You want the ability to seamlessly run commands.”
Lloyd: Warp does not store users' code, only vector embeddings
“We don't store any code. I know there's some coding tools that are actually like storing your code base. We just store vectors for semantic search.”
Lloyd: Warp uses WhisperFlow for speech-to-text API
“We're using whisper flow, which is another startup, which has got like a really good API around doing speech for text.”
Zach Lloyd: Warp relies on over one million lines of Rust code
“Warp is over a million lines of Rust code, and this is like, up until very recently, this type of agentic stuff just did not work in Warp because it's not built on any public UI framework, like we have our own internal UI framework, and so it really needs to u…”
Zach Lloyd: Warp rules differ from Cursor rules by being cloud-synced
“The biggest difference that these are is that these are sort of like cloud synced. And so you can sort of use these wherever you log in. They're sort of shareable. Across the team. And so like, this is our version of cursor's rules.”
Zach Lloyd: Warp will likely support the Cursor rule format directly
“I think we will also, by the way, probably just literally support the cursor rule format.”
Lloyd: Warp translates natural language in any REPL session
“So if you're doing like work in Node or Python, Postgres, MySQL, it works in any REPL and it's a it's a pretty differentiated and cool feature of Warp.”
Zach Lloyd: Warp's coding agent outperforms both Cursor and Windsurf
“Warp can code, and Warp code is at a level that's, like, comparable to cloud code. I think it's actually better than, like, Cursor's agent. It's probably better than Windsurf since they aren't, they don't have Sonnet for.”
Holmes: Devs will eventually rarely keep Warp open except for prompting
“And then eventually I may just check the management pane and my desktop notifications and rarely have warp open at all. Except for inputting prompts. So it feels like a more progressive interface since we own everything beyond like the CLI itself.”
Zach Lloyd: Claude Code in Ghostty cannot support editable diffs
“There's no way for cloud code running in ghosty, let's say to have an editable diff.”
Zach Lloyd: Warp's next feature will likely be in-tool agent code review
“I think honestly, like the next thing that we'll probably be building is like basically a way in warp of code reviewing what the agent is produced.”
Lloyd: Warp Will Likely Build a Headless CLI Tool for CI/CD
“I would say like our likely path forward on that is like, we will probably at some point make a warp CLI app where if you really want to run it headless, like you want to run it in CI.”
Zach Lloyd: Multi-threaded AI agents will not scale on local machines
“But if our thesis is that multi-threading is going to become really important, It's going to not be doable on a single machine. Like if you didn't like our memory usage before, you're not going to like it when we're running, you know, 10 10 agents. It's just n…”
Zach Lloyd: Cloud sandboxes face massive developer adoption friction
“I also think like the sandboxing thing is very cool, but a very big barrier to adoption in the developer tool space. We've been trying, I think, forever to get people off of their local machines to do development, whether it's like code spaces or get pod or wh…”
Zach Lloyd: Anthropic might launch a desktop agent harness like Warp
“I think even hearing the cloud code folks on your podcast, like I would not be surprised if Anthropic like launched a thing that looks a little bit more like warp where it's like a harness for running a whole bunch of different cloud codes, but it's an actual …”
Zach Lloyd: Cursor represents a very significant portion of Anthropic's revenue
“Cursor I think is some very significant portion of their revenue.”
Zach Lloyd: Warp has almost 600,000 active users
“You know, it's like almost 600,000 active users on Warp right now.”
Zach Lloyd: Warp's revenue is growing 5% to 15% week-over-week
“Our revenue since the beginning of the year is growing, like, depends on the week, between five to 15% every week, and so, like, it's actually a pretty fast revenue ramp.”
Lloyd: Warp expands Pro and Turbo request limits and adds credit top-ups
“The two plans were the pro plan and the turbo plan. Pro was at a thousand, that's going to 2500 requests. The turbo plan is going from 3000 to 10,000 requests. And then we're introducing usage based pricing.”
Lloyd: Pure usage-based pricing psychologically encourages users to ration AI tools
“The reason we haven't done the usage based, just the pure usage base is it feels psychologically like it's like in your face of like, it encourages like rationing the usage in a way that I think is, doesn't let people get the most out of it.”
Lloyd: Centralized admin controls are the primary demand from enterprise buyers
“Having the team concept also makes it easier for us to do enterprise-wide sort of like admin controls, which is honestly the, probably the biggest thing that they want is they want, like, they want to enforce that everyone's on a team, that they sign in with S…”
Lloyd: Warp cannot support on-premises enterprise configurations
“We can't support like an on-prem type configuration, which some people want.”
Holmes: Agentic Tool Calls Do Not Map Well to Go Structs
“Cause I know that more of the complaints about our backend are the fact that we leaned on Golang than the fact that we leaned on GraphQL. Which is surprising. Mostly because the way we model everything with agentic tool calls doesn't match to go structs very w…”
Zach Lloyd: Figma is entering the developer space against Lovable and Bolt
“I think Figma honestly is now like in the developer space. Like, I don't know if you all have tried any of their tools, like they're further up the like app building kind of stack, like competing with more with like lovable and bolt, but it's like every compan…”