Rajat Ujawane

Day 1 of the "12 Days of OpenAI"
ChatGPT O1 and Pro Mode are here! 💪

1) OpenAI just kicked off the 12 Days of OpenAI, and Day 1 didn’t disappoint:

✅ A new ChatGPT O1 model
✅ The introduction of Pro Mode

Let me tell you what’s cool about both 👇

Day 2 of the "12 Days of OpenAI"
Reinforcement Fine Tuning (RFT)

It’s a powerful way to customize AI for specific tasks, taking things to a whole new level. Let me explain 👇

7) It’s still in Alpha (only researchers, universities, and enterprises can apply), but it’ll be available to the public next year.

Today’s takeaway: RFT isn’t just about customizing AI, it’s about teaching it to think like an expert in YOUR field.

This feels like a huge leap.

6) OpenAI uses the same reinforcement techniques internally to train their top-tier models like GPT-4 and ChatGPT O1.

5) Here’s why RFT stands out:

✅It needs fewer examples (no massive datasets required).
✅It helps the model generalize, not just memorize.
✅It works for any domain requiring deep expertise (finance, healthcare, engineering, etc.).

4) Another wild use case: rare genetic diseases.

- Researchers fine-tuned ChatGPT to help diagnose diseases by analyzing patient symptoms and finding gene mutations.
- Instead of just spitting out guesses, the model reasoned through the problem, improving how doctors might tackle these cases.

3) Example: OpenAI worked with Thomson Reuters to create a legal assistant that handles tricky workflows.

They fine-tuned the model using just a few dozen examples.

2) Why does this matter?
With RFT, you can create expert level AI for your specific use case.

From legal assistants to scientific research, this technique allows models to excel in niche domains like never before.

1) What is RFT?
It’s not your typical fine tuning. Instead of just training the model to mimic data, RFT teaches it to reason better.

Think of it like giving the model a problem, grading its solution, and rewarding better ways of thinking.

Day 2 of the "12 Days of OpenAI"
Reinforcement Fine Tuning (RFT)

It’s a powerful way to customize AI for specific tasks, taking things to a whole new level. Let me explain 👇

6) The best part about O1?
It doesn’t just respond, it thinks before answering.
- It’s faster for easy questions.
- It slows down and gets precise for hard ones.
- It makes 34% fewer mistakes compared to the previous version.

Honestly, it just feels better to use.

5) Why would you care about Pro Mode?
Let’s say you’re into:

- Math or coding challenges
- Research-level science
- Pushing ChatGPT to do complicated workflows

Pro Mode is basically like saying Hey, take your time and give me the best answer you’ve got.

4) Now, about Pro Mode 💪

For $200/month, Pro Mode is designed for people who use ChatGPT a lot, like, a lot.

-Unlimited access (no caps or throtlng).
-O1 Pro Mode, which gives the model extra brainpower for super tough prblms.
-Advanced voice features for those who prefer to talk instead of type.

3) OpenAI showed some fun demos, including one where they uploaded a hand-drawn diagram and asked O1 to solve a problem.

The result? It figured it out in seconds.

Basically, if you’ve ever wanted an AI that can handle both words and images, this is it.

2) So, what’s the deal with ChatGPT O1?
This new version is a big step up:

- It’s way faster (no more awkward pauses when you ask it simple questions).
- It’s smarter at things like coding, math, and reasoning.

- Oh, and it can now look at images and figure things out (yep, multimodal is here).

Day 1 of the "12 Days of OpenAI"
ChatGPT O1 and Pro Mode are here! 💪

1) OpenAI just kicked off the 12 Days of OpenAI, and Day 1 didn’t disappoint:

✅ A new ChatGPT O1 model
✅ The introduction of Pro Mode

Let me tell you what’s cool about both 👇

bolt.new Prompt, run, edit & deploy web apps

Setting up a project used to take hours/days now takes minutes with tools like Bolt.new and Replit’sAI.

It’s wild how fast things are moving. Makes me wonder,are boilerplates evolving to keep up,or are they falling behind?

Either way, the future is fast 🚀

#buildinpublic

✳️ Build mental models that help you adapt across languages & stacks.

✳️ Experiment relentlessly; failure teaches faster than success.

The best engineers?
They don’t cling to tools, they master adaptability. 🚀

What do you think, let me know your thoughts 👇

#saas

The most underrated skill in tech isn’t #coding - it’s learning how to learn. 🚀

Tech evolves fast:
✳️ Frameworks you master today might be outdated tomorrow.

✳️ New tools can disrupt entire workflows.

✳️ Adopt a forever learner mindset.

✳️ Dive into the WHY behind concepts, not just the HOW.

Perfect for time-sensitive use cases like sales, events, or seasonal updates.

I’ll share more soon, but if this sounds interesting, let me know
I’d love to hear your thoughts!👇

#buildinpublic

Currently building a smart content scheduler for #webflow sites! 🚀

Here’s what it can do:
- You can schedule or reschedule any CMS item whenever you want.

- Even set your whole site to publish at a specific time.

- Want an element to appear or disappear at a certain hour? Yep, it does that too.

Perfect for time-sensitive use cases like sales, events, or seasonal updates and more.

I’ll share more soon, but if this sounds interesting.
let me know, I’d love to hear your thoughts! 👇

Could you add me?

Building in public is the best growth hack for your SaaS in 2024/25. 🚀
Here's why:
- You attract like-minded early adopters.
- Feedback loops get shorter, meaning faster iterations.
- Transparency builds trust—and trust converts.

Start small. Share wins, fails, and lessons. 🌟

✅ Stay curious. AI changes fast, and the only way to keep up is to keep learning.

AI isn’t here to take your job. It’s here to change how you work. Don’t fight it. Learn to ride the wave.

Let me tell you something I’ve learned about adapting to AI at work, it’s not about mastering every tool or trend. It’s about mindset.

Here’s what’s worked for me:
✅ Use AI to handle the boring, repetitive stuff.
✅ Pair AI’s strengths with your own: intuition, empathy, and real-world judgment.

Currently building a scheduler app ⏰designed to automate content on Webflow sites.

Think: effortless updates, streamlined workflows, and a ton of use cases for creators, marketers, and businesses 🚀

More details coming soon, stay tuned!

🦋feels different from 🧵.
It has a sense of community that 🧵 seemed to lack.

Feels like it’s shaping up to be a great space for authentic conversations.

Let’s see how it evolves! ⚡️

Entrepreneurs are the only ones who quit their 40 hours a week job to work more than 60 hours a week.

But in the end, ITS ALL WORTH IT.

bolt.new Prompt, run, edit & deploy web apps

Tried something new for my saas app landing page design

I used bolt.new to spin up a landing page with Next.js & Chakra UI.

⚡️The output wasn’t perfect, but it was shockingly close to what I envisioned.

🔥Tweaked a few things, and the final version took shape much faster than starting from scratch.