Syed Rafiq Ahmed

Project Manager at Amura Health | Ex-IPAC | Organic Growth without ads | Figuring out AI in the real world

[Syed Rafiq Ahmed @Rafiiiiq]

@launch_llama None

[Syed Rafiq Ahmed @Rafiiiiq]

@dark_coderz Execution and distribution.

Latter more important than former.

[Syed Rafiq Ahmed @Rafiiiiq]

@Abhijit_091 I already did. Outperformed a developer team of 10.

[Syed Rafiq Ahmed @Rafiiiiq]

@YashHustle_22 Think smarter

[Syed Rafiq Ahmed @Rafiiiiq]

Not getting excited, till I actually try it.

But curious, if it can just talk or also do tasks.

[Tavus @tavus]

Introducing Griffin, the first model to pass the video Turing test.

48% of people who talked to it live thought it was a real human. Previous systems have had a pass rate <3%. It is #1 on NVIDIA's benchmark for full-duplex AI video.

It’s the first Human Interaction Model (HIM).

[Tavus @tavus]

Introducing Griffin, the first model to pass the video Turing test.

48% of people who talked to it live thought it was a real human. Previous systems have had a pass rate <3%. It is #1 on NVIDIA's benchmark for full-duplex AI video.

It’s the first Human Interaction Model (HIM).

[Syed Rafiq Ahmed @Rafiiiiq]

@ai_for_success Have written my personal feedback on it.

https://t.co/RZU7miQf0I

[Syed Rafiq Ahmed @Rafiiiiq]

Honest unpaid review after using OpenAI's DOT.

It is extremely slow and we are no way close to AGI.

I used dot out of curiosity and excitment. I thought we can now finally get have our own version of JARVIS. But I got humbled. Leave AGI level stuff, Dot couldnt even do basic agentic jobs despite direct human instructions.

For example, I asked it to access one of my Google Sheets, review its contents, and write down its feedback next to each row, based on the criteria which I set.

First, It did not follow the criteria. Second, It did not have any logical reasoning for writing a certain feedback.

Third, when I asked it to give me an explanation on why it gave such a feedback, it was not able to remember what it wrote, even though that feedback was right in front of it in the sheet.

For every request or objection, I just kept hearing phrases like

"Let me look into that,"

"Let me check,"

and "I'll have to check that."

When you hit the call function, it just keeps ringing. No one picks up, and suddenly, from the background, you can hear Dot saying, "Hey, I'm listening." while the call is still ringing.

Why have a calling function when the calling function doesn't even work?

The creepiest part was when I closed the chat. Even after me clicking on the delete chat option inside Dot (which is supposed to disable the dot) it kept running. It was able to listen to me and talk to me.

I had to completely shut down the ChatGPT application and also the web application to stop it.

This is what happens when you don't test your product enough and ship just out of peer pressure.

@OpenAI I expect better from you!

[Syed Rafiq Ahmed @Rafiiiiq]

@embirico Everything.

1. Non-functional calling speaker.
2. Speed.
3. Reasoning.
4. It cant read pages which it has direct access to.

Read my post for detailed feedback:

https://t.co/RZU7miQf0I

[Syed Rafiq Ahmed @Rafiiiiq]

Honest unpaid review after using OpenAI's DOT.

It is extremely slow and we are no way close to AGI.

I used dot out of curiosity and excitment. I thought we can now finally get have our own version of JARVIS. But I got humbled. Leave AGI level stuff, Dot couldnt even do basic agentic jobs despite direct human instructions.

For example, I asked it to access one of my Google Sheets, review its contents, and write down its feedback next to each row, based on the criteria which I set.

First, It did not follow the criteria. Second, It did not have any logical reasoning for writing a certain feedback.

Third, when I asked it to give me an explanation on why it gave such a feedback, it was not able to remember what it wrote, even though that feedback was right in front of it in the sheet.

For every request or objection, I just kept hearing phrases like

"Let me look into that,"

"Let me check,"

and "I'll have to check that."

When you hit the call function, it just keeps ringing. No one picks up, and suddenly, from the background, you can hear Dot saying, "Hey, I'm listening." while the call is still ringing.

Why have a calling function when the calling function doesn't even work?

The creepiest part was when I closed the chat. Even after me clicking on the delete chat option inside Dot (which is supposed to disable the dot) it kept running. It was able to listen to me and talk to me.

I had to completely shut down the ChatGPT application and also the web application to stop it.

This is what happens when you don't test your product enough and ship just out of peer pressure.

@OpenAI I expect better from you!

[Syed Rafiq Ahmed @Rafiiiiq]

Honest unpaid review after using OpenAI's DOT.

It is extremely slow and we are no way close to AGI.

I used dot out of curiosity and excitment. I thought we can now finally get have our own version of JARVIS. But I got humbled. Leave AGI level stuff, Dot couldnt even do basic agentic jobs despite direct human instructions.

For example, I asked it to access one of my Google Sheets, review its contents, and write down its feedback next to each row, based on the criteria which I set.

First, It did not follow the criteria. Second, It did not have any logical reasoning for writing a certain feedback.

Third, when I asked it to give me an explanation on why it gave such a feedback, it was not able to remember what it wrote, even though that feedback was right in front of it in the sheet.

For every request or objection, I just kept hearing phrases like

"Let me look into that,"

"Let me check,"

and "I'll have to check that."

When you hit the call function, it just keeps ringing. No one picks up, and suddenly, from the background, you can hear Dot saying, "Hey, I'm listening." while the call is still ringing.

Why have a calling function when the calling function doesn't even work?

The creepiest part was when I closed the chat. Even after me clicking on the delete chat option inside Dot (which is supposed to disable the dot) it kept running. It was able to listen to me and talk to me.

I had to completely shut down the ChatGPT application and also the web application to stop it.

This is what happens when you don't test your product enough and ship just out of peer pressure.

@OpenAI I expect better from you!

[Syed Rafiq Ahmed @Rafiiiiq]

I had two options: learn first, or learn while building.

I haven’t posted here in over a month because I was building an AI product. I also wanted to learn product management properly.

One option was to finish a certification before starting. The other was to learn while building, applying each concept to the product as I went.

I chose the second.

Before starting the build, I worked with ChatGPT, using Astra, to lay out a structured learning and development plan. Much like a course, it had a sequence. Each stage connected what I needed to understand with something I needed to do.

We started with product discovery: understanding the problem, who faced it and how they handled it already.

Then came experiments to check whether the idea could deliver value, including benchmarking AI models for accuracy, speed and cost. From there, we moved into defining the product, deciding what belonged in the first version, and building and testing it.

Whenever I got stuck, AI explained the relevant product management principle through examples, then helped me apply it to the decision in front of me.

I still had to use my own judgement. AI’s guidance wasn’t always right, so I had to question its reasoning and correct the direction when needed.

By the end, I had a working AI MVP and practical experience with concepts I had previously only read about.

Here are a few ways you could try this:

• Learn product management: build a support chatbot.
• Learn storytelling: script and shoot a short video.
• Learn automation: build a weekly reporting workflow.

What helped me was having a plan to follow, a project to apply it to, and feedback as I went.

If there’s a skill you’ve been meaning to learn, choose a small project that needs it. Work with AI to map out what you need to learn and how to apply it to that project.

You don’t have to know everything before you begin. Start building.

[Kathryn @kathrynwu1]

Being smart is overrated in startups. Everyone in places like YC or Silicon Valley is smart, but it doesn’t move the needle.

What actually pushes companies forward is not intelligence, but being resourceful, clear, and brutally honest with yourself.

The best founders I've seen own their mistakes fast and pivot hard.

Smartness is a baseline. Clarity and grit are the real bottlenecks.

[Syed Rafiq Ahmed @Rafiiiiq]

Payment failed but order still went through.

Just contacted customer service to get a payment mode as I was not physically present at my home.

Even after 10 mins of wait, no one attended the chat.

Meanwhile, delivery agent arrived at home.

P.s. Ironical part is, the description of the chat box says “One block customers are our top priority “.

If this is how top priority customers are attended, I wonder how non one block customers are treated.

Time to leave Swiggy!

[Syed Rafiq Ahmed @Rafiiiiq]

@officialepfo @epfochennaizo @rochennaisouth This issue has been resolved. Thank you.

[Syed Rafiq Ahmed @Rafiiiiq]

@khushiirl thousands of tokens? A heavy user using codex can easily burn 50-100 million tokens daily.

My max is 135 million tokens in a day.

[Syed Rafiq Ahmed @Rafiiiiq]

@ChrisUniverse @0xJinu @ChatGPT @thsottiaux Same, I bought a pro, was at 99% remaining weekly usage when I got a reset to 100%.

A banked reset could have been more useful.

[Syed Rafiq Ahmed @Rafiiiiq]

@codex_resets is it for all paid users or just pro?

[Syed Rafiq Ahmed @Rafiiiiq]

@officialepfo @epfochennaizo @rochennaisouth any update? @officialepfo @rochennaisouth @epfochennaizo

[Syed Rafiq Ahmed @Rafiiiiq]

@statusisdown Yes