Whole world is competing on AI benchmarks with their own foundational LLM models. Meanwhile people in India be like holy shit a well funded startup just integrated video generation api with WhatsApp in 30 lines of codes.

Kinda wanna train an ai model based on the messages I have sent on discord till now just to see how it behaves.

Damn it

If I finetune it hard, it keeps making cringey generations. Life huh. But hopefully I eventually figure out something.

Started playing with my own X scraper yesterday cuz I was bored with my main projects. But now I’m bored with this too. Burn out is real.

Damn it, this freakin code keeps getting stuck in loop. Sorry for the spam again.

Almost got cooked last night for a critical bug in Cleus. I should stop testing updates in production ffs.

60000 users complete on Cleus, and 40000 more to go for to reach this year's target.

Chrome is so slow after the latest update. It keeps freezing.

Vibe coders with no previous experience with programming and how security is handled be like.

Talking to Meta AI in Rayban glasses kinda feel like talking to Siri, I wish there was an option switch to Gemini or GPT like models.

Integrated GPT-5 to my discord bot. But I noticed most responses for questions kinda starts with “Short answer:”. Wasn’t happening before. Not an issue tho, but was little unusually repetitive.

Tryna make an uncensored version of OpenAI's GPT OSS 20b model. Let's see how it goes.

The GPT-5 hate kinda feels forced.

Feels weird to see Cleus in the same conversation as GPT-5.

OpenAI's GPT OSS 120b is so censored that it won't even tell a lie for fun.

Just made Cleus AI much much smarter but at the cost of speed.

I’m so cooked, I saw myself fine tuning an ai model in my dream tonight.

Is it just me or ChatGPT is responding at a speed of one token per second right now. I’m on plus subscription too.