Vikas Tiwari

agentic ai product guy

[Vikas Tiwari @pmvikastiwari]

Let's pin this forever!

[Vikas Tiwari @pmvikastiwari]

@JayneshKasliwal Drink less water

[Vikas Tiwari @pmvikastiwari]

I see this exact message atleast once every day on X. I wonder how does it still get traction.

[Lachha Paratha @usernameonXs]

Twitter is cool.

But it’s 100x better when your timeline is full with people who code and build things.

I need to connect with more founders and tech people. If you’re into Tech, AI, Startups, Design, web dev, SaaS, ML, App development or programming, say hi

[Xiangyi Li @xdotli]

sharing my personal library on evals 1/n

i put together the highest quality blogs, podcasts, papers, and projects on evals. additions are welcome!

[Vikas Tiwari @pmvikastiwari]

DREAM SO BIG THAT SOMETIMES IT IS HARD TO BELIEVE.

NEVER EVER GIVE UP!

ALWAYS KNOW THAT THERE'S MORE PEOPLE THAN YOU CAN NAME WHO IS LIVING YOUR DREAMS.

[Vikas Tiwari @pmvikastiwari]

It's hard to look at DMs if they want something but can't give you anything in return.

Always remember this when you DM someone and they don't look at it. They're not a bad person, you don't provide them enough value.

[Vikas Tiwari @pmvikastiwari]

@benhylak Are u looking for a product person?

[Vikas Tiwari @pmvikastiwari]

Acknowledged that my mental health is deteriorating, and went to a professional for help, then came back, and completed the work.

[Dear Son. @DearS_o_n]

As a man, during your darkest period, what was the best thing you ever did for your mental health? Please be HONEST.

[Dear Son. @DearS_o_n]

Men can stay up til 2 a.m., wake up at 6, be in debt, broke, alone, and still have faith that one day, everything will work out. It's called being a man.

[Vikas Tiwari @pmvikastiwari]

😂😂😂

[Vikas Tiwari @pmvikastiwari]

Agents are amazing. You just need 5 things to understand everything:

1. Agent.md - Instruction for agents. Can be many as well, based on the task and architecture
2. Agent Skills - Skills.md, description, and instructions with reference if required
3. MCP - Agents to access tools and external data sources
4. A2A - Agent-to-Agent communication via Agent Card, which is nothing but an agent description for the task
5. Subagents - Agents spawn for multi-parallel and serial costs for better task allocation, context length management, and also to become adversarial validators.

Everything else is just additional layers to solve your problems, but agents in all it's fundamental sense doesn't need anything else.

[Vikas Tiwari @pmvikastiwari]

Biggest challenge for any Indian builder is to make senior citizen cohort above 60-65yrs, pay or invest any kind of money in their apps.

For now, the only value builders get from this cohort is DAU, and it helps them buy more time through findings.

I wonder if Kunal Shah can change it through WhatsApp, I don't know if my parents would spend money to use it.

[Vivek Raju @vivekraju93]

Few thoughts on CRED/Shah/Meta/Whatsapp

1. If Kunal Shah left CRED for Whatsapp without Meta putting in a significant sum in CRED, it would have looked like jumping a sinking ship. Shah must have pushed for the investment as part of the deal, especially with IPO plans looming (not immediate but in a few years).

2. Not sure how Meta and CRED fit together. Could be a pure financial investment to get Shah. CRED's entire model is attracting as much txn volume as possible from the creditworthy and earning different amounts of spread on them (PG, bill pay will be nil to low spread, lending and lifestyle will have higher spread). Whatsapp or any of Meta's products don't organically fit into Cred and vice-versa. Atleast as of now.

3. While Silicon Valley insiders know of Shah, the media side of the Valley are not very aware of him. It is going to take a while for them to learn about him and his prowess. Expect a few deep profiles on him (and probably hit pieces as well if someone has it out for Zuck).

4. CRED just had its first profitable quarter. Very high chance that this is not a blip but a trend that will continue. The reason - Shah lost 35 kgs. Don't laugh yet. Let me explain. A common trend seen among loss-making tech startups is their founders being to looksmaxx around the time profitability kicks in. Amazon, Airbnb, Zomato, Meta, Zerodha, Groww, Lokal. Very strong correlation. Of course it's not causation. Probably an outcome of knowing the ship is stable now, and having more time & peace in their mind to focus on themselves.

[Vikas Tiwari @pmvikastiwari]

If you're his manager, then you were not approachable enough. I inferred it by your tone that you were actually surprised, and that only happens when you don't know about them.

Talking helps resolves 99.99% of the things, so you don't have to come on X screen-posting how bad you look.

[Vikas Tiwari @pmvikastiwari]

Some of the biggest problem with Gemini app, which is affecting it's adoption over Claude and GPT is:

1. Even though context length is "1 Million tokens", the output is often far lesser and trivial, and user have to ask again and again, and run through multi-conversation thread to achieve what they want. It doesn't follow prompt instructions well.
2. Native Flowchart, HTML support: It still asks user to download mermaid code and view it somewhere else; same goes with HTML and other native easily solvable problems. Massive disadvantage if I have to go on different tools to get answered
3. No Skills support in native app. Atleast allow me to chat with Gems.
4. I can't chat with Youtube videos, it sucks. Infact, others have done far better job than Google, despite owning the ecosystem. Sucks so much that they removed support completely.
5. It can't write emails, it's the most basic ones to automate. I hope sparks solve this problem but haven't tested yet.

I don't know if it's a solve-all mindset that is clouding their brains in solving this simplest gaps, but something is missing. I don't think Google is inept to do this.

[Vikas Tiwari @pmvikastiwari]

It takes real guts!!!!

[shyamal @shyamalanadkat]

after close to four years at @openai, i moved from the bay area to india earlier this year. i still believe deeply in ensuring true superintelligence accelerates science and remains accessible and beneficial to all. having grown up here, i've also always felt deeply connected to the ecosystem here.

over the past several weeks, i've been speaking with researchers, engineers, and thinkers across india and apac. it's become clear that there are many who want to build the future from here. moving back felt like the counterintuitive choice. i no longer think that's true.

what's been missing is the belief that you can build institutions of global consequence from anywhere. and more importantly, the ambition and the will to pursue ideas that seem impossibly large at first. this may be a once in a generation opportunity.

more to come soon. DMs open if this resonates.

[Vikas Tiwari @pmvikastiwari]

@shyamalanadkat @OpenAI All power to you man! Less goo

[Vikas Tiwari @pmvikastiwari]

Google Chat did this way back, but unfortunately no one uses Google Chat.

[Claude @claudeai]

Introducing Claude Tag, a new way for teams to work with Claude.

In Slack, Claude joins as a team member with access to the channels and tools you choose. Tag Claude in and delegate tasks to it while you focus on other work.

[Vikas Tiwari @pmvikastiwari]

The release of GLM 5.2 is a massive signal that the era of 'gated' intelligence is ending. Attempting to restrict global access to frontier-level models is a losing battle. Innovation can’t be bottled up by policy or proprietary walls.
This is just the beginning; the next generation of open-weight models is already on the way to outpacing the status quo.

[Baidu AI @BaiduAI_News]

We’re open-sourcing Unlimited OCR — built to read long documents in one pass.

With 3B total parameters and only 500M activated, Unlimited OCR sets new end-to-end SOTA results on OmniDocBench v1.5 and v1.6.

The key innovation is Reference Sliding Window Attention (R-SWA), inspired by how humans transcribe books: keeping the source, recent context, and next words in focus, while softly forgetting what’s no longer needed.

With constant KV Cache size and lower attention cost, Unlimited OCR can transcribe 40+ pages in a single forward pass — without losing context or slowing down.

Explore the model👇:
--GitHub: https://t.co/5ZJBsEldKd
--Hugging Face: https://t.co/4FKFr9EfOu