The analogy I keep coming back to is photography after cameras got cheap. The title photographer didn't disappear, it just changed. It went from the person who can operate complex equipment and develop film to the person with taste, vision, and the ability to make a specific image happen, and it's the same thing with software engineering. The technical floor collapsed but the artistic ceiling got higher
AI models are not like football teams, I have no loyalty to any company and I'll use them based on how much value they give me.
my first LLM experience was Jurassic-1 from AI21, an Israeli company whose model was close in capabilities to GPT 3, which I couldn't access yet because of openai's long waitlist. I switched to GPT 3 the moment they gave me API access and that's when I knew this was the future.
then GPT 3.5 was a huge leap. it felt weird to me when they launched ChatGPT because I didn't really see a difference, and I kept using the API since it was cheaper and gave me more control, but who would have thought that the interface would change everything.
GPT 4 was the first time I actually thought AGI is really possible with these models. around the same time, the smartest AI to just chat with, even if not for anything useful, was HeyPi (the company microsoft didn't really buy but ended up hiring all the founders and most of the team), and it had something amazing I didn't see in GPT 4, which was access to the latest information, but GPT caught up soon enough.
I tried Claude 2 and it sucked, but everything changed again with Claude 3. this AI was so good to talk to that I remember being in Vegas playing poker, and instead of focusing on the cards I had Claude on my phone chatting with it all night. it felt human and smart, and that's what pushed me to make a 1h30 video just talking with it, which ended up getting over 400k views on youtube.
I kept Claude as my main AI until o1 preview came out. I still think at that point Claude was the most humane and the most emotionally smart model, but o1 preview showed me something else entirely, like how freaking smart it was compared to anything in the world. it was the first AI that made me feel humans were soon going to be outpaced.
Soon after, openai launched GPT Pro for 200 dollars a month and it was the best investment tbh, and even though it's been over a year and a half and 3600 dollars in payments to openai, I can count at least 20x the value I got from it directly and probably 50x indirectly. I still kept my Claude subscription because I liked talking to it and it was good for programming and for cursor.
for the last year this didn't really change, I still thought openai was at the top with its models and nothing would beat their Pro models in most things, but for programming Claude was my go to when I needed help coding quickly. that changed with GPT 5, GPT 5 codex, and especially 5.4 and 5.5. GPT became so good at coding that even though it still lacks in front end, there are ways to make it good, like using gpt image 2 alongside it. Claude is still useful, but GPT 5.5 can now do almost everything, and especially when paired with codex it understands you very clearly and implements your vision without throwing in all the extra things Claude used to add.
one small thing is that as much as I love GPT 5.5 on codex, GPT 5.5 Pro doesn't show as much value anymore. I feel like it was nerfed hard, and I still believe old GPT 5.4 Pro had way more value than this one even if 5.5 Pro is quicker.
during all this time I didn't mention other models because Gemini is promising but it's so lacking at following instructions, Deepseek had its hype moment but it just sucks, and I lost hope for Grok because with all those massive data centers it still couldn't make a name for itself beyond being an edgy AI fine tuned on right wing stuff.
Conclusion for now, I feel like openai is on top. they have general use locked up with GPT 5.5, and now they added coding on top of that. it's hard for anthropic or google to compete, especially with how anthropic is treating its users and its developers.
what really gets me about anthropic is the whole holier than thou act they keep putting on, like they're the responsible adults in the room saving humanity from dangerous AI, but if you actually look at what they're doing, they're going hard after open source AI models and trying to shape the laws to make sure they can kill them, all while hiding behind this idea of we care about safety and preventing harm narrative. and Mythos is the perfect example of this, because in practice GPT 5.5 Cyber and even GPT 5.5 Pro are right there at the same benchmark level and sometimes even slightly ahead, so that whole too dangerous to release theater was basically marketing. Honestly at this point anthropic feels more like a cult than a real AI lab tbh.
so for now I'm team openai, but if any other lab releases the next level AI I'm switching for sure.
فالمغرب إذا كنتي إنسان معافى وسليم نفسيا مغاديش يزوجوك واليديك ،إذا كنتي مريض نفسيا كيكون عندهم الحاح كبير باش يزوجوك باش يتهناو منك ويجيبو ليك شي ضحية من النسوة اللواتي أنهكهن سؤال المجتمع : واش باقي ما جاب الله؟
لن تجد أسودا يدين هذه الأعمال ،إن تضامنهم فيما بينهم حتى فيما يتعلق بسلوك الزومبي هو عقدة نقص ،جرب أن تضع هذا البشر عند حده وسترى البكائيات وتهم العنصرية.
⭕ توضيح : لفظ اسود ليس عنصريا نقول اسود لأنه أسود بالفعل ومن لا يعجبه لونه فهو عنصري في حقه نفسه.
معمر كانت الصحة والتعليم أولوية، لو كانت لما خرج الشباب اليوم يطالبون بهما بعد 69 سنة من "الفرص المهدرة".
حنا في 2025 و8 وفيات في شهر واحد في مستشفى عام بسبب الاهمال وسيدات حوامل ينقلن على النعوش.
مستشفيات عبارة عن جدران فقط.. لا اطباء لا تجهيزات ولا معاملة آدمية.
#مونديال_الصحة