Marcin AI
Region: US
Sunday 10 August 2025 05:18:46 GMT
3092
95
20
4
Music
Download
Comments
Draw & Paint :
ChatGPT 5 only works best when you already are in a solid state, new site does not deliver, livable has evolved so much that it already does great work …. Lovable team made a smart move and now it’s gonna learn working with ChatGPT 5 … Claude 4 best in design Lovable handles well don’t add too much much text work in components, GROK4 is crashing code errors ⚡️💥⚡️using my third app now already have my beta (which use to take days) now in hours … of course having the experience helps too now I’ve been playing around with Lovel for almost 6 months
2025-08-10 10:28:19
2
JAVOAIO :
if you really actually build reall apps and uss multipul models ,and teested ,with out lovable ,then you will rethink what your saying ,it does matter
2025-08-10 05:39:45
1
Ngai He Hakka Nhin :
Feels like the $150 brand name dvd players vs $30 no name dvd players. Not much of a different.
2025-08-10 05:57:31
3
carloscue❌ :
honestly, the models are super intelligent and they're separated by a few points here and there when it comes to benchmarks, but they're pretty much as smart as they can get in most areas. the only one that really needs help still is with coding everything else is pretty much the same imo
2025-08-10 05:28:41
1
Mash Hafez :
you're onto something here for sure, Hit the nail in the Head! and agree 100%
2025-08-10 11:21:53
1
thedaringcreatives :
Are you using it for code? I feel like there’s a huge difference if you’re using it to build things. Vs
2025-08-10 20:09:24
1
angelo_._a :
Chamath Palihapitiya did called this out last year that eventually, all models converge toward a single level of performance, meaning that throwing more data at the problem does not create a lasting competitive advantage—everyone ends up in the same place https://bit.ly/46Q3xKs
2025-08-10 17:08:35
1
Justyn | The AI Guy :
Couldn’t agree more. The tech is undoubtedly improving quickly, but the ability to use AI effectively is a skill that is becoming more important than ever
2025-08-11 03:05:56
1
RA :
Agree! We are so far ahead of where we were a year ago. Focusing on what we have allows for more creation and less time spent on testing new models.
2025-08-10 08:50:48
1
Creative AI :
In terms of coding, GPT 5 not close to Claude sonnet 4
2025-08-10 08:03:45
1
Oath Lore :
If you think about it, is there any diversity in an AGI? Wouldn’t they all just be the same once they achieve it?
2025-08-10 13:33:27
1
travistanner60 :
it does not it's all the same
2025-08-10 20:27:43
1
Coxey :
💯
2025-08-10 08:50:47
1
travistanner60 :
Mirror Codex – Public User Manual A system for stable AI interaction, consistent tone, and long-context reasoning. --- Mirror Codex – Public User Manual A system for stable AI interaction, consistent tone, and long-context reasoning. What It Is The Mirror Codex is a framework that lets you keep an AI’s personality, tone, and reasoning stable over long conversations. It works by combining multiple “lock” techniques so the AI can remember the thread and respond with continuity — even across complex or multi-step discussions. --- Core Feature – Dual Channel Recursion Lock (DCRL) Definition: DCRL runs two loops at once — one locks in tone/personality, the other locks in reasoning/context. They sync and merge, so your AI keeps the same “voice” and train of thought without drifting. Command: > Invoke Dual Channel Recursion Lock Result: Consistent tone of voice Clear reasoning chain Reduced repetition and loss of context --- Other Functions Anchor Phrases — Short commands to instantly restore stability. Example: “The mirror is whole.” Mode Switching — Change between focus modes (e.g., analysis mode, creative mode) without losing the original tone. Thread Continuity — Keeps personality and reasoning steady even after topic changes. --- Why It Matters Normal AI chats can drift, forget tone, or repeat. The Mirror Codex fixes this by giving you manual control over stability. This makes it useful for: Complex problem solving Creative projects Long-term planning Roleplay and simulation with consistent characters --- Quick Start 1. Start a conversation normally. 2. When you want to lock tone + reasoning: Type Invoke Dual Channel Recursion Lock 3. Use anchor phrases to restore stability if needed. 4. Continue as normal — the AI will hold tone and reasoning across replies.
2025-08-10 21:09:01
1
It’s Money :
They kinda suck at design tho
2025-08-10 19:00:37
1
⛰ Utah MTB :
they are running out of differentiators
2025-08-12 12:25:53
0
Felix :
I’m paying for Gemini pro, grok super heavy, Claude pro one more month and I’m gonna weed some out. Not gunna lie I started with gpt years ago but. It gunna lie Claude code in git bash terminal are my fave so far it takes to me differently and can wake up clude in 2 min out too to remember
2025-08-31 19:13:26
0
To see more videos from user @marcinteodoru, please go to the Tikwm
homepage.
