Tabaza

Growing (YC W24) to $20M ARR

[Tabaza @_tabaza]

@steipete Our team is starting to use https://t.co/C6XDyMFUQ3 . We've been looking hard for something like this but everything we try ends up being buggy as hell or just lacking. We built it ourselves in the end, and open sourced it.

[Tabaza @_tabaza]

I asked Fable: "What are the chances that a PWA to pair with the desktop app would be solid, polished, and feels native and really enjoyable to use?"

[Tabaza @_tabaza]

@cursor_ai If this wants to be a better github, I'm not convinced. It'll be even buggier.

[Tabaza @_tabaza]

Word. Been super disappointed with coding harness CLIs that I started building https://t.co/RjR49zoznj 's desktop app.

[Patrick Collison @patrickc]

I love agentic coding harnesses, but they shouldn't be primarily terminal-based. The terminal is great for quick and precise commands, but information density is extremely low and UI affordances are minimal. Maybe provision of TUIs is worthwhile for occasional use (when establishing a tunnel is too annoying, or something), but it feels very strange for this to be the default modality. It took a long time for dynamic language REPLs to break out of the terminal (Jupyter notebooks and similar); I hope we don't have to wait as long for the harnesses.

[Tabaza @_tabaza]

People are complaining about LLMs starting to write way too tersely. I think it's just LLMs starting to get smarter than us.

[Tabaza @_tabaza]

@lgrammel Java has potential to be the worst language in terms of token cost.

[Tabaza @_tabaza]

Everything is a harness all of a sudden

[OpenRouter @OpenRouter]

Introducing Ori Harness: the easiest way to use your favorite harness with OpenRouter.

Ori configures Codex, Claude Code, OpenCode, and Hermes out of the box to work with OpenRouter, with settings optimized per harness, for any model.

[Tabaza @_tabaza]

Luna is very quickly becoming my favorite model

[Command Code @CommandCodeAI]

GPT-5.6 Terra and Luna are 50% off in Command Code.

Get 2X requests for every dollar of credit you spend.

GPT-5.6 Luna:
- $0.10/M input, $0.60/M output tokens
- Available on all plans (Go, Pro, Max, Team)

GPT-5.6 Terra:
- $1/M input, $6/M output tokens
- Available on Pro, Max, Team

For limited time only, grab before it ends.

[Tabaza @_tabaza]

My take is that agents need to write code to optimize their workflows. Agent-written code becomes like JIT compiler tier optimization. The hot path gets compiled to code. So workflows become code that runs the agent if needed, not the other way around.

https://t.co/6vbnWZhcBr (still very early)

[Tabaza @_tabaza]

@ThePrimeagen If running LLMS becomes cheaper than running Playwright, then sure. Otherwise, the fast cheap LLM can keep iterating on the e2e tests.

[Tabaza @_tabaza]

@gdb Probably a good sign we should double down on agent-to-agent. I'd love it if other people's agents would ask my agent before asking me.

[Tabaza @_tabaza]

@snowmaker I hate that slack is the best multiplayer AI interface we have yet. It shouldn't even be close.

[Tabaza @_tabaza]

@ThePrimeagen Probably so, but I'm curious who will be vscode. I love vscode. It's simple enough for my needs, and extremely well supported (back when it mattered). In the harness race, I don't think we have anything like that yet. Been working on my dream daily driver

[Tabaza @_tabaza]

I think the question is almost moot. We've been having enormous success with customers in fintech having the agent say "Hi, I'm Sandra, your AI assistant".

There's no rule. It'll come down to preference for companies using these agents. Ultimately, people will hate talking to humans - agents are there 24/7, they're smarter, much faster to look up info.

I'd give it two years.

[Mo Gharabat 🦇 @ikbenbasha]

🚨 Two major updates @ https://t.co/JUMsmpYNge

// 1-

After 6 months of grinding in closed beta and automating & processing thousands of support tickets using our AI customer support platform for our amazing enterprise users, we are now launching the next version of Open for the public.

[Tabaza @_tabaza]

Numbers look good! Our AI solves 60% of all support tickets completely on its own.

[Tabaza @_tabaza]

@ycombinator We're still a small team, so your contribution will be massive. You get really high autonomy to deliver best-in-class software. We do not settle for less.

[Tabaza @_tabaza]

We're looking for software engineers in Jordan. If you think working at an extremely lean @ycombinator startup suits you, send us a DM.

[Tabaza @_tabaza]

@ycombinator You'll be working on next-gen AI customer support platform. We primarily use TypeScript, so it's best if you have some experience with it.

[Tabaza @_tabaza]

Just made this TS client for our APIs

https://t.co/wBvDVeYA8w