Aug 5, 2025 · 37m · latent-space
⚡️OpenCode: Claude Code but Open Source, with Any Model, and frontier TUI - with Dax Reed (@thdxr)
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
Dax Reed joins Latent Space to present OpenCode, an open-source, model-agnostic terminal coding assistant built with a full-screen TUI, modular architecture, and enduring developer ergonomics to rival proprietary tools like Claude Code.
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 24.5% of the talking time here. How this is scored →
The hosts as informed peer 5.6 Guest teaching 4.6 Guest disagreement 2.7 The hosts pushing back 2.1
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 25:38 Stop submitting unsolicited PRs
Dax forcefully tells the community to stop sending uncoordinated PRs and argues that open source by committee generally makes products worse.
Hardest push from the hosts ▶ 17:11 Swyx challenges dismissal of compaction limitsSwyx pushes back against Dax's minimization of context window compaction issues, citing power user complaints and proposing non-destructive conversation RAG.
Biggest teaching moment ▶ 17:37 Dax on the obsolescence of clever model hacksDax educates Swyx on the danger of premature optimization in AI tooling, explaining how specialized fast-apply models were completely wiped out by Sonnet 3.5/4.
The host holds their own ▶ 30:59 Swyx analyzes Cursor's defensive acquisitionSwyx demonstrates sharp industry expertise by breaking down how Boris and Kat built the primary existential threat to Cursor, forcing Cursor to acqui-hire them.
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 |
|---|---|---|---|---|---|---|
| Introductions and Terminal-First Experiments | 5 | 3 | 2 | 1 | Hosts introduce Dax Reed and inquire about OpenCode's origins and divergence from Claude Code. Dax sets the framing around developer UX rather than chasing marginal prompt-engineering gains. | |
| OpenCode TUI Demo and Session Sharing | 5 | 4 | 1 | 1 | Dax demos the fullscreen TUI interface, diff views, and session-sharing URLs. Alessio notes distinct features like session snapshotting, which Dax explains was created for enterprise PR review workflows. | |
| Avoiding the Embedded Code Editor Trap | 6 | 4 | 2 | 2 | Swyx draws parallels to Warp's struggle with building an embedded code editor. Dax strongly validates this, explaining that building an editor is an infinite rabbit hole and clarifying that agent loops are simple tool loops without magic. | |
| Technical Architecture: Bun and Client-Server Design | 6 | 7 | 4 | 4 | Dax corrects Swyx regarding SQLite and details his Bun single-file executable architecture. When Swyx presses on context window limits and suggests non-destructive RAG, Dax reframes it, arguing that fancy prompt hacks are quickly obsoleted by frontier model improvements. | |
| Practical Benchmarking and Tooling Pitfalls | 5 | 5 | 3 | 2 | Dax dismisses synthetic coding benchmarks like mazes as irrelevant to real-world software engineering. He shares a candid story about introducing regressions by blindly adopting Gemini CLI fallback strategies. | |
| Enterprise Telemetry Conservatism and Token Economics | 5 | 4 | 2 | 1 | Alessio queries model usage and token economics, noting suspected losses at Anthropic. Dax explains OpenCode's strict zero-telemetry default for enterprise adoption and argues flat-rate plans are loss-leading growth subsidies. | |
| Open Source Strategy and Ecosystem Positioning | 7 | 5 | 5 | 4 | Dax delivers contrarian takes on open-source product design and firmly requests contributors stop submitting uncoordinated PRs. Swyx and Dax engage in strategic analysis of Cursor acquiring the Claude Code team to mitigate an existential threat. |
Statements from this episode (23)
Dax Reed cofounded a thriving coffee company requiring SSH to purchase.
“About a year ago, we launched a coffee company me and a few other people. And the trick was you can't buy it off a website. You have to SSH into a server and you actually get a really good experience. Like it's actually good UX and it's a pretty good e-commerc…”
Dax Reed: Optimizing marginal LLM efficiency yields plateauing returns.
“I think the clarity we have is we're not trying to get the next three percent of efficiency out of the LLM. I think that's kind of like an infinite rabbit hole you can dive down that has like plateauing results. So for us, we're more focused on Product experie…”
Dax Reed says working with AI is code review, not code writing.
“Like when you're talking to these LLMs, the work you're doing is not writing code, it's reviewing code.”
Dax Reed: Adding code editors to terminal AI tools is a trap.
“Almost immediately we're like, oh, we should build edit functionality. Like we should let you edit files in here, but that's a trap because that is like an infinite rabbit hole of work. And I've seen this with warp already where I think they have an editor in …”
OpenCode exactly replicated Claude Code by dumping its system prompts and schemas.
“We look at cloud code. We dump all their system prompts. We dump all their tool descriptions. We dump all their tool schemas. We re-implement the tools. And when you're using an anthropic model, we basically have the exact same implementation.”
Dax Reed says AI agent harnesses are fundamentally simple tool-execution loops.
“At the end of the day, like it's just a loop where you send the prompt with a bunch of tool descriptions. It tells you what tools to call. You call it tools and the results back. And there's really not too much magic there.”
OpenCode uses extensible custom modes rather than a single fixed plan mode.
“We're probably gonna ship it this week, but we're not implementing it as An explicit plan mode. We're implementing the concept of modes. So you'll be able to hit tab and cycle through different modes. We'll have a built in like plan mode and implement mode, bu…”
OpenCode uses a decoupled client-server architecture unlike Anthropic's Claude Code.
“So the one other difference between us and cloud code is under the hood, there's a client and server architecture here. So open code is a server that has a built in TUI client.”
The OpenCode team is developing mobile, web, and desktop clients.
“We're starting work on a mobile client. We're gonna start work on a web client. We're gonna have a desktop app.”
Anthropic's Claude Code job posting specifically seeks engineers with Bun experience.
“I've seen now cloud code in their job posting. They actually mentioned specifically they're hiring someone that has experience in bun executables.”
Dax Reed notes developers avoid context limits via frequent session restarts.
“To be honest, most people haven't complained about this. Like it comes up occasionally. But people tend to start new sessions pretty frequently. They tend to not have super long running things that often. So it is a problem that needs to be solved, but it's on…”
Dax Reed says complex optimizations on top of LLMs quickly become obsolete.
“Any kind of fancy optimization you're doing on top of the LM eventually gets, like, you kind of lose the need to do it given how fast things are moving.”
Dax Reed says almost no AI coding benchmarks resemble real-world engineering tasks.
“Yeah, so we looked at all the different benchmarks we could find, and it's almost comical how none of them look like my day-to-day work. Most of these evals are like, You know, solve this maze. And I'm just like, I'm never solving a maze. Like it's never anyth…”
Dax Reed claims the Gemini CLI codebase is rushed and corrupts edits.
“And Gemini ones were not good. It was initially it looked good, but then I noticed it was getting caught in these crazy loops and it was like editing my files in these totally messed up ways. And I should have known because you can tell when you're looking thr…”
OpenCode will offload terms-of-service-questionable integrations to community-maintained plugins.
“I expected that open code is gonna have a plugin system in the next couple of weeks. So we'll probably move all of those like TOS questionable integrations into community maintained plugins and people can do whatever they want there.”
Dax Reed says enterprise developers immediately reject tools collecting code telemetry.
“Open code fits in a weird space where there's a lot of very boring enterprise companies that haven't used any of these tools yet because they have certain restrictions in their corporate laptop where they can't even install them. Open code tends to fit nicer i…”
Dax Reed predicts Claude's $200/month pricing is an unsustainable growth strategy.
“I think it's pretty easy to, like, use more than 200 dollars worth, even by accident. So I would also lean towards that the Claude Max plans are a growth strategy, not like any long term pricing thing that can work, at least at the current, given the current s…”
Dax Reed argues open-source products are usually worse due to committee design.
“We don't think a product is good because it's open source. Usually a product is bad because it's open source. It tends to be like designed by committee and like just a smash of like a bunch of different competing priorities.”
Open source excels at core architectures needing hundreds of long-tail integrations.
“The stuff that has like, there's a core. And once you put the core, there's like a hundred integrations to do. That's where open source really shines. And there's a lot of that in open code. That's why we made it open source.”
Dax Reed predicts OpenCode will dominate when a competitor beats Claude Sonnet.
“What would change things is if there's a day where either another LM lab or like, you know, an open source model drops that is competitive with Sonnet, maybe even better than Sonnet on that day, open code is going to be the only way to do this kind of thing. C…”
Dax Reed predicts Cursor will move faster than Anthropic's Claude Code team.
“I mean, when I saw the acquisition, I was like, oh, this is actually more intense now because the cursor team is going to move faster than the cloud code team. I think that's my feeling given they have to, there's more pressure on them and they're smaller.”