levi

🥷

[levi @nagylevidotdev]

@jacobAlswin wanderair

[levi @nagylevidotdev]

@thsottiaux reset for the weekend?

[levi @nagylevidotdev]

@manxlab The first impressions (marketing page -login - onboarding - app home page) must be polished, those are the most important to hook a new user. Then proceed shipping new features. People tend to polish everything, you must not.

[levi @nagylevidotdev]

@appllamaio @jaimintf Subscribed for 1 year, very good deal. Let's get building

[levi @nagylevidotdev]

@bidah or just ask the AI friend to implement it in fastlane brother 5-10 minute to create a proper pipeline. EAS has advantage over OTA + release process, and ofc cert handling and such (which are the cherry on the top - not the deployment)

[levi @nagylevidotdev]

@marclou analysis paralysis. I still think that stupid people has a bigger rate to succes in general.

[Tobi @tobiadonadon_]

i built a skill to un-slop AI-looking apps

most AI-generated UIs still have the same problem:

they look polished, but they also look obviously AI-made

same cards
same gradients
same generic “modern app” layout
same zero-personality design

so this skill forces claude to actually think differently than it would normally do

→ starts from a real visual concept
→ explores 3 different directions
→ picks one coherent design language
→ renders the screens at iPhone size
→ scans the result for obvious AI-slop patterns
→ critiques and refines it before you ship

when everyone is shipping ai slop apps, yours has to be different to succeed

i’m giving the full skill away for free

rt + comment “stop slop” and i’ll send it over
(follow so i can dm)

[levi @nagylevidotdev]

@tobiadonadon_ stop slop (let me check that it is usable in expo too - not just figma slop)

[levi @nagylevidotdev]

The skill that changed how I start to develop a feature: https://t.co/TnsDv37mBm . Creating new stuff is not the hard part, but making it usable and actual is the key to success.

[levi @nagylevidotdev]

@Kappaemme1926 I have an M4 Pro with 48GB of RAM.. it is barely enough when running multiple subagents in codex and iOS Simulator.. With that in mind, if you are going with a lower-spec Mac, dont forget to upgrade the storage in it.. 512GB is not enough at all ,especially with swapping.

[Tibo @thsottiaux]

Hi Astra users. A reset and a quick update on quality issues that have been posted around.

Working with some of you, we have found and fixed the following issues:
- Some skills written for previous models were triggering too often or preventing the model from checking its work.
- An opt-in context management experiment that could cause early stops or replies to older messages. We've disabled it. Our rough estimate is that 4-5k users were affected by this experiment.
- We've also removed some badly configured engines that resulted in a measured quality degradation for a long tail of traffic flowing through them.

We’ve also made some more minor improvements and things should feel significantly better across the board. More consistent follow-through, better tracking of your latest message, and better checks on the work as it’s going through the motions.

The examples posted and all the users who worked directly with us were incredibly useful in helping fix things quickly. Always grateful for this incredible community.

And of course, a reset is also landing by midnight today.

[levi @nagylevidotdev]

@Saanvi_dhillon terminal

[levi @nagylevidotdev]

@jackfriks @postbridge_ and 1B token / day

[levi @nagylevidotdev]

day 4

wtf am i doing

[levi @nagylevidotdev]

@jp54362 definitely better

[levi @nagylevidotdev]

@melvindvivas I do believe pro 5x must be enough for high-throughput but optimized workflows. I have been working on multiple projects concurrently and it is kinda hard to run out of usage after the optimizations I have done. no context context engineering burn tokens regardless of plans.

[levi @nagylevidotdev]

@melvindvivas kinda better, but it mainly because of enforcing 5.6-sol-high orchestration with 5.6-luna-max subagents and 5.6-sol-med validator. outside of coding, it shows improvements. I am abusing it on 3 concurrent projects + UI rendering with gpt-image-2 and all good atm.

[levi @nagylevidotdev]

@melvindvivas terra-high has less slop than luna-max in coding tasks. It really matters if you prompt your orchestrator well then luna-max slop can be handled by 5.6-sol-high orchestrator. I mostly use terra-high if I don't need an orchestrator and I dont wanna write code.

[levi @nagylevidotdev]

@jonaheapen98 whoops

[Klaas @forgebitz]

i am somewhat of a prompt engineer myself