Offscreen

Apps, websites, and side projects built quietly. Useful sometimes. Interesting, hopefully.

[Offscreen @QuietlyByG]

You’re not behind on AI models.
You’re behind on invoices.

Same day this week, builders were posting their stack tax:
• SuperGrok / Grok plan: $30/mo — “fast” burned ~$13 of API plus a weekly-limit hit
• Claude Code: $100/mo
• Codex: $100/mo (already cancelled)
• Grok Bot: $30/mo
• Hermes: free, and some people route it under one coding sub to dodge agent API bills

The winning move isn’t collecting every new model.
It’s keeping one paid coding harness and cutting the rest.

[Offscreen @QuietlyByG]

Codex CLI just got sticky chat headers so the prompt doesn’t disappear while you scroll the run.

That’s the kind of change that actually cuts friction: you stop losing the thread mid-task, not “we shipped a bigger model.”

Same lesson as shipping apps — half the pain is the loop around the work (scroll, context, where you left off), not the headline feature.

If you live in agent CLIs: what’s the smallest UX fix that saved you more time than a model bump?

[Offscreen @QuietlyByG]

@NivTheNitro Same setup here. Release branch is whatever App Store Connect will actually take. Duo stuff stays in a side branch until Xcode catches up. I only pull over the bits that compile clean on stable.

How are you deciding what counts as “safe to ship” vs “park it”?

[Offscreen @QuietlyByG]

iPhone Duo APIs ship in the iOS 27.1 SDK.
App Store Connect won't take beta builds for review.
Stable Xcode can't compile the Duo code.

So you keep two branches, or you wait.
That gap is the actual release plan.

[Offscreen @QuietlyByG]

@Shadowfetch Yeah this is the move. I now treat the first successful archive + TestFlight upload as a milestone, not something I do at the end. Once that’s working, the actual feature work feels way less stressful.

[Offscreen @QuietlyByG]

Half of "shipping an iOS app" is waiting. Certificates. Review. That one Xcode error that only happens on the archive.

Notes from actually shipping to the App Store (not a tutorial):

You can write the feature in a day. Then you lose three days to signing, provisioning, and "works on my phone, fails in TestFlight."

App Store Connect has its own weather. Metadata ready, screenshots right size, privacy labels filled, then you wait. Sometimes the blocker isn't code. It's a missing export compliance answer or a screenshot that cropped wrong on iPad.

What helped me: ship a thin vertical first. One job. TestFlight with real devices early. Keep a checklist so Connect doesn't become archaeology the night before submit.

If you're mid-build: what's eating more time right now, Xcode or Connect?

[Offscreen @QuietlyByG]

@MarioNawfal Says the corrupt Dumb fck

[Offscreen @QuietlyByG]

@MarioNawfal His Son is a billionaire… fckn hypocrite

[Offscreen @QuietlyByG]

@revistaoeste Hahahahahaha

[Offscreen @QuietlyByG]

@oalanicolas Grok build para mim é uma excelente ferramenta para desenvolvimento de softwares , trading bots e indicadores , o astra e o claude nao conseguem entender o que eu quero mas o grok build é excelente nesse tipo de tarefa

[Offscreen @QuietlyByG]

Astra 6 low eating a ton of tokens on simple tasks.

Anyone else seeing this?

Low still burns reasoning tokens you never see, then the agent rereads files, calls tools, and “double-checks” a one-line change.

Is that expected, or is the harness overthinking everything?

What’s actually working for you — tighter prompts, smaller scope, or just dropping back to Sol/Luna for the easy stuff?

[Offscreen @QuietlyByG]

@SidnelsonEu Hahahahaha

[Offscreen @QuietlyByG]

@KoroushAK Cool , i am testing an EA that trades breakout only for gold , and the backtest and the demo live results are very promising

[Offscreen @QuietlyByG]

Fair point — and I agree.

This wasn't meant to be a controlled Low vs High benchmark; it's a real-world comparison of two workflows I actually shipped.

A proper follow-up would hold the same app, spec, repo and acceptance tests constant, change only reasoning effort, and measure:

retries to pass
source/research quality
edit distance from first draft to accepted build
latency
token/usage cost
final defects

I’d also track how many issues each run catches before I point them out.

That would be a much cleaner Astra Low vs High test. Might be the next experiment.

[Offscreen @QuietlyByG]

https://t.co/O9Naa7nlzU

[Offscreen @QuietlyByG]

@videoclaw only mac avaliable for now ? any launch date for windows?

[Offscreen @QuietlyByG]

Louco demais dahora parabens !
To no meu 8o ios app , 2 softwares de trading e alguns bots que tinha feito com o Sol 5.6 , agora com astra ta corrigindo a par de problemas de bug q eu nem sabia que tinha mas drena igual eu no bar kkk…
Se tiver mais projetos maneiros posta ai pra galera acompanhar

[Offscreen @QuietlyByG]

@witcheer cool fixed some stuff on my hermes!

[Offscreen @QuietlyByG]

@mfriasoficial Nuremberg para todos os cumplices do sistema!