wintercounter

Front-end specialized full-stack engineer with Open Source passion and controversial takes.

[wintercounter @wintercounter1]

@glass_awful @theo Highly doubt. Used up $50 worth of credits in just 3 prompts in 20 minutes, and that was back with 4.7.

[wintercounter @wintercounter1]

@stewartcelani @BHolmesDev But most of the dev envs require *nix. Especially if you use Docker.

[wintercounter @wintercounter1]

Sure, with subscriptions, once you need to use an API, it's fucked. Just to quote from Claude's own docs about PR reviews: Each review averages $15-25 in cost, scaling with PR size, codebase complexity.

What you do for $150 in a month using subs would burn the same amount of credits in 30 minutes using API...

[wintercounter @wintercounter1]

@markmcerqueira @kentcdodds The reason is with batching 1 slow query will block all others.

[wintercounter @wintercounter1]

@jxnlco @rosslazer * do NOT show up...

[wintercounter @wintercounter1]

@jxnlco @rosslazer Doesn't work well, sessions do show up on the phone, and it needs to run a separate server. Claude doesn't need it, I can enable it permanently or mid session.

[wintercounter @wintercounter1]

Being a UI for my codex inside WSL seamlessly, using my own config from home instead of needing config copying/syncing and env var hacks.

Working remote control, no sessions show up on my phone at all, and I shouldn't need to run a separate server for it, I should be able to turn it on permanently or in the middle of a session.

[wintercounter @wintercounter1]

@theo I just moved from Codex to it. Since 5.5 my weekly limit is reached in 3 days ($200 sub), I had no choice. Since SpaceXAI Claude limits got better than Codex, I have no choice.

[wintercounter @wintercounter1]

@KaiXCreator Photonic circuits.

[wintercounter @wintercounter1]

@yashhq_22 FYI Claude has better limits now than Codex since Elon.

[wintercounter @wintercounter1]

@sama Limits, after 6 months of Codex I switched to CC because since GPT 5.5 I use my weekly limit in 3 days on the Pro 20x plan. I tried adding extra credits to keep using, $40 1000 credits, it was enough for another 3 prompts...

[wintercounter @wintercounter1]

@DanielSmidstrup Cuz it's the perfect drug. Thankfully Codex weekly limits help me disconnect after 3 days using GPT 5.5 on 20x Pro plan...

[wintercounter @wintercounter1]

@DanIsBuilding @claudeai Are they really competitors?! It's like comparing Paint to Photoshop.

[wintercounter @wintercounter1]

@MatthewBerman Try again with impeccable, it helps a lot on design.

[wintercounter @wintercounter1]

@GergelyOrosz Well, there goes my $20 sub, I only used it as a design reviewer for Codex.

[wintercounter @wintercounter1]

@dsog @chrisgpt ■ {"type":"error","status":400,"error":{"type":"invalid_request_error","message":"The 'gpt-5.5' model is
not supported when using Codex with a ChatGPT account."}}

[wintercounter @wintercounter1]

@yol_kin @kettanaito And here's the problem with it. Your first request's response needs to opt-in first so that subsequent requests will send it. So you still cannot use it on first visit, rendering this header completely useless.

[wintercounter @wintercounter1]

@scott_butler_jr @kentcdodds Search is possible also within their own UI.

[wintercounter @wintercounter1]

@marcel_butucea @Cointelegraph We need to have e2e encryption. Only the user and the provider should be able to read any input/output tokens.