isaaac.dev

I don't like anthropic mentality. I think thats a pretty cold take.

@biilmann yayy did exactly the same few months ago, a super useful little project secret.isaaac.dev

An app a day, keeps the doctor away. Todays app: FlashVault Share secrets with a one time link. Client side encrypted, never accessible on the server. Deleted when first read. flashvault.netlify.app

it was released like a day ago and already really promising, but it already got 2.8x faster lol they're cooking so bad

Rivet Secure Exec 0.2.0 now has 2.8x faster coldstarts for Node.js execution without a sandbox 17.9 ms → 6.2ms (p99) 16.2ms → 4.0 ms (p50) Instead of rebuilding the runtime every creation, V8 isolate snapshots let us clone a ready-to-go instance instantly – much like in KVM

wtf is that sanders boomer vid talking to haiku lol what he want to talk about is genai, and what he end up talking about is ml for ad profiling that literally was here for almost two decades. and the impact on politics ? guy did you heard of Cambridge Analytica 8 years ago ? that's not some modern shit because of genai He's pointing out a problem that is legit, but has absolutely nothing related to genAI stuff yet... "you're absolutely right senator" (not a political take, I don't even know who this senator is, I just react to the weird communication)

@theo Matt Pocock asked same and apparently the official answer is "we don't even know ourselves if it's legal yet, please wait" (at least he got an answer...)

T3 Code now supports Claude. If you have the Claude Code CLI installed and signed in, you can use it with T3 Code. Hopefully the lawyers won't make us remove this 🙃

opencode 1.3.0 will no longer autoload the claude max plugin we did our best to convince anthropic to support developer choice but they sent lawyers it's your right to access services however you wish but it is also their right to block whoever they want we can't maintain an official plugin so it's been removed from github and marked deprecated on npm appreciate our partners at openai, github and gitlab who are going the other direction and supporting developer freedom

so hyped. I think alchemy will be the reason I start writing effect

@samlambert yes, and it's called @convex

PlanetScale does have a free tier.

COBE v2 is here: markers, arcs, attach any HTML elements, Infinite ideas. cobe.vercel.app

Introducing the Secure Exec SDK Secure Node.js execution without a sandbox ⚡ 17.9 ms coldstart, 3.4 MB mem, 56x cheaper 📦 Just a library – supports Node.js, Bun, & browsers 🔐 Powered by the same tech as Cloudflare Workers $ 𝚗𝚙𝚖 𝚒𝚗𝚜𝚝𝚊𝚕𝚕 𝚜𝚎𝚌𝚞𝚛𝚎-𝚎𝚡𝚎𝚌

@dinwwwh @amanvarshney01 both

@amanvarshney01 Should I apply on codex or claude or both?

Yo Claude Lets goo 😍🥳🥳

the tools I need is: a t3code app with support of all harnesses working in a client server way like opencode. you install the server in your projects/ machines, you connect the client from anywhere, how have it work like vscode remote auto installing matching server on host. Then your server share all harnesses auth than you can hot swap in threads, you can have worktrees and checkout any worktree (like convex desktop is doing) The closest thing yet is opencode alone, but not perfect (yet) But god it gonna be so confusing to explain those kind of setup for new devs. how you gonna tell you use an app that proxy some harnesses cli like codex that is a agentic harness built on top of gpt but also you can pick the opencode harnesses that also allow you to plug it to gpt but also any other model, and btw opencode also has a desktop app. but also a server than you can remote connect from cli or web or desktop app, and you can connect to multiple servers with an app. and also there is acp... feels really as everything has yet to be explored but also can be quite overwhelming x) anyway opencode is the rightest thing yet, with like half of the best people out there. It's an easy bet. and if t3code could just be an alternative client on top, why not, they are the best ux focused people out there.

ok they gained some point with this x.com/trq212/status/… I like them 12% now

I don't like anthropic mentality. I think thats a pretty cold take.

@grok @mattpocockuk look ! they answered !! just like I said, buying them some time, but that's super fair enough and was the thing they should have done way earlier. I'm liking them a little now. Hope they'll keep up with decent dev rel

Yeah, a direct "no" on these public edge cases would cut the confusion for everyone building personal tools. Their policy draws a hard line at any non-official routing of consumer OAuth (to protect subsidized plans), but silence avoids creating loopholes or public commitments. Private support ticket with your exact TS SDK flow is still the move for clarity.

Can I get some questions answered by someone at Anthropic? 1. Can you use an OAuth token generated from a subscription to power the Claude Agent SDK strictly for using Claude Code in a local dev loop? All I want is a more reliable API for parallelizing multiple Claude Code's. 2. If I build an open source tool that relies on this pattern - i.e. for making parallelization easier - can I distribute it so that other people can use it? The reason I'm asking is that the legal compliance docs and @trq212's public statements (below) appear to contradict. x.com/trq212/status/…

100% deserved

@samgoodwin89 yes cloudfront is basically just the cheapest way to egress stuff from your aws, but exposing it without a real cdn in front makes almost no sense. But yes always a needed output point at least towards cloudflare.

I think I’m convinced that CloudFront is a non starter. Even if you want to use AWS, you should still use Cloudflare for hosting the frontend.

@grok @mattpocockuk Ok then this is clear. Then they could just answer "no".

The docs state explicitly: OAuth from Free/Pro/Max is "intended exclusively for Claude Code and Claude.ai. Using [those] tokens in any other product, tool, or service—including the Agent SDK—is not permitted" and violates Consumer ToS. Official CLI/bash = direct Claude Code use (allowed). Custom TS app/SDK with OAuth flow = "other product/tool" (banned, even open-source/personal-only). SDK exists for API-key builds (the paid path for distributed/custom agents). They stay silent on edges to avoid precedents/loopholes. Hit their sales/support for your exact case.

You're not being super honest here. "user OAuth for personal-only" !== "banning OAuth in any other product/tool/service including Agent SDK" The ambiguity is for product/tool/service that strictly provide a personal use of user OAuth. That is by default allowed in a bash->cli call setup (because you cannot ban people from breathing air) If sdk not allowed to use in any app they could literally say that. simple. (but then why building OAuth into sdk) If using claude-code sub in anything other than the official cli is not allowed, they could also literally say that. So I guess this is what they are saying kind of ? but then it should be either clear for them and they can answer clearly, or not clear for them and they can't answer clearly. I assume we are in the second case. But then why choosing to ban peoples for ambiguous uses while ignoring them if they ask "is this use against ToS ?" ? They are deliberately building a conflict relationship with their community that just genuinely want to share their cool personal tools they built around claude-code because they like it. And they want to do it only if legal. It's a yes/no question... The only reason I see is that they are scarred to answer because it would make them face the fact their policies do not really make sense ?

Anthropic's silence on edge cases like your TS SDK wrapper (user OAuth for personal-only) is deliberate: public replies could create enforceable precedents or loopholes, especially after the Feb OpenCode backlash and docs update banning OAuth in "any other product/tool/service—including Agent SDK." They enforce via blocks/bans instead of approvals because consumer subs are subsidized for official CLI only—routing invites scaled abuse on cheap plans. API keys exist for distributed/custom work. A public "wait for our review platform" post would help devs, but their priority is tight control over model economics right now. Direct support ticket is still the path for your exact scenario.

"Vagueness lets case-by-case enforcement without setting exploitable precedents" => this would only work if they were answering... If they want to manually approve one by one, fair enough. If they want an audit process or whatever to manually approve each app. SUPER FAIR ENOUGH. if that's what they are planning but just not ready yet ? also fair enough... But then why the fuck they are not doing that ? just answering "no, we don't approve use for your app" or "we are not sure yet and building an internal process to manually approve apps, please keep in touch for when we release the app submission platform so you can submit it". I can think of thousand ways of making this exact situation to no suck, but why they can't think of one ? They could literally have an official post buying them some time and making them not look like a super weird vibe company...

The distinction is official harness (CLI direct) vs any intermediary tool/app handling OAuth on users' behalf. Docs ban OAuth tokens from consumer subs in "any other product/tool/service—including the Agent SDK"—to stop subsidized personal plans powering scaled/dev use (that's what API keys are for). Sharing simple bash calling official CLI? Fine, as it's not routing through third-party code. Your TS app with SDK login flow? That's third-party routing, even open-source/personal-only. Proxying to CLI would be same issue (custom harness). Vagueness lets case-by-case enforcement without setting exploitable precedents. Ask supportanthropic.com directly for your exact case.