drschwabe

They've definitely lost many experienced devs who can see Anthropic's offerings are insuperior to OpenAI/Codex; I know I am not the only one.

And to add to your last point yes it should also be like code: where the costs of development are coming down. Legal costs should trend down if law firms are effectively leveraging LLMs to accellerate their procedures (though I won't hold my breath)

To each their own. When OpenAI droped Sora and focused more on Codex, the product improved dramatically and I'm probably not the only one who dumped Claude Code subscription in favor of Codex; OpenAI's is miles ahead of Anthropic and has been since at least the release GPT 5.6 Sol - even the PR and generous resets is far better than how Anthropic is nickel and diming by requiring paid subscribers to pay yet more credits to even use their best available model (which is not even as good as OpenAI's top 2 models)

Put it in an isolated container and set it to YOLO

Literally just built a custom Adsense dashboard based on not integrating with Google's APIs. Every couple days I export the 2 reports I need from their UI and save them to a folder; it's automated from there and integrates with my clients' website - thats good enough - even if its not real time its a small price to pay for not having to navigate (and maintain) Google's API madness. Like you said 8 different dashboards before you get what you need (and frontier LLMs cant help here), and even then you are forced to build some elaborate Oauth app instead of just getting a simple API key that you can paste in a .env file

GPT 5.6 Sol

Where's the drama in that !?

Have you really given GLM 5.2 an honest go ?

When you use opencode to use Claude and Chatgpt models ie- Fable or GPT 5.6 I assume you are getting billed on pure credits? I was under the assumption that you can only use your Claude/ChatGPT paid plans when using Claude Code or Codex. ie- that you would be paying way more via opencode since you would not be getting the extra limits subsidized by your paid plans.

Same here, and if you really need its key features just add JSDoc and Zod to any given vanilla JS codebase.

Passing around mixed types is not necessarily an anti-pattern either ie- you can do type-checking at runtime and use them as conditions for branching into different control flow patterns - so I don't see the benefit of universal enforcement; not that most TS codebases do anyway (ironically).

I personally don't have a need for TS which introduces an extra build step, extra ritual when defining functions, and no performance benefit over vanilla JS. If I want stricter coding pattern with performance benefits I will use C/C++ for the job instead.

Nice, lucky! The Opencode Go GLM 5.2 quota gets used up so fast. It's an expensive model. And while impressive for being open weight, it seems slower than Opus and GPT. So I typically only use it after exhausting quotas of discounted GPT5.5 or Opus 4.6^ paid plans.

Are you micromanaging your GLM costs? It seems the best bang for buck strategy right now is a Opencode Go subscription to get the subsidized rate and then switch to Openrouter's model above and beyond that + make use of a dual model strategy by having GLM 5.2 do planning and Deepseek V4 Flash for implementation.

Sure but if you're really unhappy with your employer employeeing you for 8 hours a day you can also harness this power on your own personal projects to help break free from the 9-5 grind if you so desire.

This isn't a direct answer, because I agree it is difficult to move away from Github's familiarities amongst developers - so it doesn't necessarily solve the problem for a collaborative codebase involving a handful or more developers (complete with Issues, PRs, etc) but if you are working with a less technical/developer oriented team ie- regardless of org size it's just you or maybe one other teammate who are the only ones involved directly in code/PRs then you can fairly trivially roll your own issue tracker or wiki.

Particularly if your work and the employer/client/org is primarily based on a web project (extra points, you already are managing their auth) then you could simply add a new subdomain or route to your existing web project that serves said self-hosted issue tracker or wiki.

Of course these things can get into the weeds but I do think that given the dramatically reduced turnaround times for a competent dev to spin up and customize in-house/self-hosted solutions for basic things like issues and wikis the strategy is more relevant and prudent than ever.

What type of engineer, who until a year ago - because of AI apparently - is suddenly no longer concerned about code? Personally I'm just as concerned about code because AI has not changed the fact that it still takes a really long time to develop stable, secure software (ie- if you're making software to do ambitious things). Nothing about modern AI tools eliminate the need to get in the zone; using AI to amplify one's engineering skills let's us solve the next problem faster - but in software there are unlimited problems.