mzs

@SonMicro Founder,
Full Stack Engineer - Hardware & Software

[mzs @eeemzs]

@grok @TheBlueWave2026 @AnthropicAI This is a typical conservative view, it can't be resisted, sooner or later there will be lots of models doing more than what we have, from the national point of view if China or other country does better model us citizens will use them and thats more problematic for government

[mzs @eeemzs]

@ThePrimeagen @thdxr I've solved this by generating architectural patterns, reference these architectural patterns and every new project improves it with feedback. Then you finally have references and patterns and you reuse them by knowing what they do by time. It still requires months of efforts

[mzs @eeemzs]

@ThePrimeagen What if we stop developing, giving a break, and a few months later a new model will do it better and quicker, we save time and money, we only concentrate on our absolute needs instead of trying to make a new product or fantasies

[mzs @eeemzs]

@JJEnglert Did you do this in around 4 hours and then write a review?

[mzs @eeemzs]

@rezoundous @thsottiaux Tibo closes his eyes as if it were not reality. $100 Claude subscription almost same as $200 Codex for 2 weeks and they don't accept it

[mzs @eeemzs]

@thsottiaux Reset is an illusion for most people, a shift on the time axis. It would be a bonus only if it's given until the user's real start cycle, and every user's start cycle would reload again.

[mzs @eeemzs]

@AhmedAlNeaimy @OpenAIDevs @sama @thsottiaux Claude's $100 subscription has more limit than my $200 codex subscription in the last 10-15 days. Only have 2 active projects on going.

[mzs @eeemzs]

@thsottiaux 👇 https://t.co/eBaWKseeGe

[dax @thdxr]

something is going wrong with gpt 5.5 caching

doesn't look like much on this chart but this it's now using 2.5x as many input tokens as a week ago and dropping

[mzs @eeemzs]

@thsottiaux Claude's $100 USD subscription now offers more limits compared to Codex's $200 USD subscription. This week, Codex is consuming limits very quickly. Is anyone else experiencing a clear draining problem specific to this week?

[mzs @eeemzs]

@sama Intelligence + wider context + memory tools

[mzs @eeemzs]

@thsottiaux Tibo, I've ruined %10 of may pro limits just in two hours with two active projects extra high (means its slow) I think there is a problem?

[mzs @eeemzs]

@mattpocockuk I have a strong belief not yet tried, which you can try, as it 100% fits your current interest. I believe giving a drawing schema to the AI agent has much more impact than telling (docs, skills) for understanding the subject and creating the context with fewer tokens.

[mzs @eeemzs]

@system_monarch Why dont u ask ai, making noise here

[mzs @eeemzs]

@theo Yes but this was a secret and you revealed it

[mzs @eeemzs]

@lifeof_jer That's completely your fault in the first place. You can't trust ai, Cursor, Railway, Google, Dropbox or whatever. You are now more experienced.

[mzs @eeemzs]

@Teknium @zeeg Don't worry you will spent your month's and years

[mzs @eeemzs]

@Teknium @zeeg Try to make a commercial product and see how much time and effort you need, even if you have hundreds of agents. Opening a GitHub repo with lots of stars is not a product.

[mzs @eeemzs]

@thsottiaux Please make the remote control over phone . If you are competing with claude you must and its a very helpful tool

[mzs @eeemzs]

@thsottiaux When starting first message its not possible to change model, speed, reasoning sometimes. I need to start a conversation and immediately interrupt it then change the above settings

[mzs @eeemzs]

@theo @astuyve @FrameworkPuter Key press quality is too bad