https://teapawt.net/t.co/MRJ6Mz7aOV Building, learning, engineering and writing
Nagpur, India
Joined February 2023
- Tweets29.6K
- Following777
- Followers1.7K
- Likes27K
Tanisha Pritha retweeted
Okay, tell me one thing.
If your AI agent fails when the API returns a 429, is it really autonomous?
Okay, tell me one thing.
If your AI agent fails when the API returns a 429, is it really autonomous?
Tanisha Pritha retweeted
AI made building cheaper.
Why does getting your first 10 customers still feel so hard?
Tanisha Pritha retweeted
What does a 70% coding-agent benchmark score actually tell you?
A. The model can code
B. The model can solve complex tasks
C. The agent is production-ready
D. Not enough information
Tanisha Pritha retweeted
They're running out of names at this point 😭
Introducing Gemini 4 Carbon, our latest frontier model.
Carbon delivers improved performance across domains — ranking #1 on v5.0 of the @ArtificialAnlys Intelligence Index and sets a new state-of-the-art in cybersecurity and long-horizon capabilities.
It’s rolling out starting this week to cyber defenders through our Fairwind Program, with expanded access coming soon 🧵
Tanisha Pritha retweeted
A. [[1], [], []]
B. [[1], [1], [1]]
C. [1, 1, 1]
D. TypeError
What's the output? 👀
They're running out of names at this point 😭
Introducing Gemini 4 Carbon, our latest frontier model.
Carbon delivers improved performance across domains — ranking #1 on v5.0 of the @ArtificialAnlys Intelligence Index and sets a new state-of-the-art in cybersecurity and long-horizon capabilities.
It’s rolling out starting this week to cyber defenders through our Fairwind Program, with expanded access coming soon 🧵
What does a 70% coding-agent benchmark score actually tell you?
A. The model can code
B. The model can solve complex tasks
C. The agent is production-ready
D. Not enough information
Tanisha Pritha retweeted
If vibe coding is really the future…
Where are the vibe-coded
- programming languages
- compilers
- operating systems
- databases
- browsers
- payment systems
- security protocols
We trust AI to build apps.
But do we trust it to build the things those apps depend on?
Maybe the real bottleneck isn't code generation.
It's trust.
If vibe coding is really the future…
Where are the vibe-coded
- programming languages
- compilers
- operating systems
- databases
- browsers
- payment systems
- security protocols
We trust AI to build apps.
But do we trust it to build the things those apps depend on?
Maybe the real bottleneck isn't code generation.
It's trust.
An AI agent demo can look magical.
Then you test it with:
- A tool timeout
- An empty response
- A duplicate request
- A failed database write
- A user changing their mind halfway through
Suddenly, the real engineering begins.
Reliability is the product. The LLM is just one component.
Tanisha Pritha retweeted
You can build a genuinely good product and still get 0 users.
Before writing another feature, try this:
1. Find 20 people with the problem
2. Talk to 5 of them
3. Learn where they already spend time
4. Write about the problem publicly
5. Share one useful solution
6. Get 3 people to try it
7. Watch where they get stuck
8. Repeat before adding features
Distribution starts before launch.
Not after.
You can build a genuinely good product and still get 0 users.
Before writing another feature, try this:
1. Find 20 people with the problem
2. Talk to 5 of them
3. Learn where they already spend time
4. Write about the problem publicly
5. Share one useful solution
6. Get 3 people to try it
7. Watch where they get stuck
8. Repeat before adding features
Distribution starts before launch.
Not after.
Tanisha Pritha retweeted
Interview question:
Your agent calls a payment API.
The API processes the payment, but the request times out before the agent receives the response.
What should the agent do next?
A) Retry immediately
B) Tell the user it failed
C) Check the transaction status using an idempotency key
D) Ask the LLM to decide