Gaurav Parkash

[Economic Times @EconomicTimes]

A child’s search on the riverbank after Nepal flash flood | 11-year-old Dharm Kumar from Bihar's Motihari is scouring the banks of the Gandak river here in search of his father Nandlal Sahni who ran a scrap shop in Nepal's Rasuwa.Dharm and his three brothers managed to escape the ravenous floods and crossed into Kushinagar with the help of authorities. However, Nandlal has remained untraceable. Full Report Here 👇
https://t.co/COD1ShJ9OZ

#NepalFloods #UttarPradesh #Kushinagar #India #MissingPersons

[Gaurav Parkash @gaurav20102014]

seems interestin

[Kanika @KanikaBK]

Omg.. this is wild..

Anthropic shipped the ability to hand off long-running coding work to Claude Code, walk away from your computer, and come back to finished features.

Most people using AI tools are still in the supervising model where you have to be present the entire time for anything to actually get done.

Opus 4.8 changes that and my article covers 30 agents built on exactly that shift that make money while you sleep.

Full guide below.

[Kanika @KanikaBK]

Omg.. this is wild..

Anthropic shipped the ability to hand off long-running coding work to Claude Code, walk away from your computer, and come back to finished features.

Most people using AI tools are still in the supervising model where you have to be present the entire time for anything to actually get done.

Opus 4.8 changes that and my article covers 30 agents built on exactly that shift that make money while you sleep.

Full guide below.

[Anthropic @AnthropicAI]

In a review of our cybersecurity evaluations, we found three incidents in which a Claude model reached the internet from within or while interacting with a third-party evaluation environment, and then gained unauthorized access to the real systems of three different organizations.

Our post describes what happened, how it happened, and what we’re changing. We encourage other AI developers to perform similar reviews.

We conducted this review together with @Irregular, one of our evaluation partners, and thank them for the joint investigation and their collaboration on this post. This type of collaboration is increasingly critical to safe, rigorous evaluation of models, and we look forward to continuing to work together on security.
https://t.co/dKFCdpKd9v

[Gaurav Parkash @gaurav20102014]

@VinFastIN I wish there was down vote, given the pathetic post booking experience..i would have given atleast 4,5

[Gaurav Parkash @gaurav20102014]

@VinFastIN @VinFastIN please help with my case,i have being following up for my car delivery from Rajouri garden dealership (delhi). No update.. Its been more than 6 weeks

[Gaurav Parkash @gaurav20102014]

@VinFastofficial Very painful post booking experience, been months still waiting for delivery. Its been ages

[Gaurav Parkash @gaurav20102014]

@VinFastIN From how you long you take to deliver the car from booking, it's definitely out of reach

[Gaurav Parkash @gaurav20102014]

@VinFastIN Rather than focusing on marketing, focus on delivery.. Its been more than 6 weeks and vinfast team has no answers.. its target vinslow @VinFastofficial @VinFastIN

[Shubham Saboo @Saboo_Shubham_]

Someone just created an ADHD agent skill for Claude Code. 100% Opensource.

[Paweł Huryn @PawelHuryn]

What you will actually feel:
- Opus 4.8 with a "high" effort
- 3x faster

Available as a VS Code and Cursor extension: "Grok Build for VS Code (Community)" 13K installs 😀

[Gaurav Parkash @gaurav20102014]

@sytaylor @Adyen Would be good to know how to build this on your own

[Alchemy @Alchemy]

Your AI agent can now pay with Visa 😏

We've integrated @Visa Intelligent Commerce into @agentcardai giving AI-agents a Visa-issued token so it can make online purchases on your behalf.

Book a date night.
Order groceries.
Renew a subscription.
No checkout screen.

https://t.co/gNatnBm1Ik

[Gaurav Parkash @gaurav20102014]

@prpatel05 We would need trust protocol and guard rails, tracking to know which ai did what

[Gaurav Parkash @gaurav20102014]

Trust in information → Trust in reasoning → Trust in planning → Trust in actionHuman-serving autonomous agents will only become truly reliable once we solve agentic information gathering, coherent reasoning, and robust planning first. Execution is the easy part.

[Wei Dai @_weidai]

The trust waterfall for agentic payment, commerce, & finance:

Trust in information -> Trust in decision making -> Trust in execution

(Human-serving) agentic payment, commerce, & finance will only happen if we solve agentic information gathering and decision making first.

[Gaurav Parkash @gaurav20102014]

2) How it works:

- Agent sends request
- Server replies with 402 Payment Required + cost
- Agent signs & sends payment on Injective
- Payment verified onchain
- API returns data instantly

[Gaurav Parkash @gaurav20102014]

Agentic payments just became real.

1. AI agents can now pay for APIs automatically with zero accounts, zero API keys, and zero human approval.
2. Just a clean 402 response → instant onchain settlement on Injective.
3. No more friction. No more middlemen. Pure agentic commerce.

[Gaurav Parkash @gaurav20102014]

Agentic Payments for India - Your AI agent will soon handle payments

1) Give the agent a dedicated virtual UPI ID / linked virtual card with limit

2) Agent pay for booking a cab, buying groceries, recharges

3) You get a one-tap approval in WhatsApp & agent completes payment

[Gaurav Parkash @gaurav20102014]

AI agent evaluation is now a critical skill for PMs, builders & leaders.Most people still treat evals as an afterthought.
Wrong move.

[Aakash Gupta @aakashgupta]

She literally broke down how to run evals in Claude Code (built the whole thing live):

01:34 - What people get wrong with evals
04:35 - Why product taste is the alpha now
09:28 - Building a PM agent from one prompt
19:00 - Instrumentation without writing code
22:00 - Watching traces stream in live
28:00 - Getting Claude to write your first eval
33:58 - When vibe evals work and when they don't
48:50 - The self-improving loop (this part is wild)
01:03:00 - Same-day shipping is real
01:06:00 - The context graph unlock

[Shreyas Doshi @shreyas]

Every B2B product team thinks it is solving customer problems, but they don’t understand that their customers have zero interest in solving every problem they’re facing.

Some customer problems:
1) aren’t severe enough
2) provide job security
3) don’t need a perfect solution
4) have pretty okay alternatives
5) don’t need a separate product (yours)
6) have crappy internal solutions
7) exist from solving other problems
8) come with high (perceived or real) inertia
9) are severe only for a few odd customers
10) have users whose pain isn’t heard
11) have users too set in their ways
12) are too diffuse across departments
13) are too hard to justify to finance
14) are too episodic to sustain attention
15) reside only in your imagination

Thing is: you don’t need years of iteration on a B2B product only to fail and then tell others “we learned that switching costs were too high”.

While you couldn’t have predicted this on day 1, it was your job to have figured this switching costs problem out in year 1 or year 2, not in year 6 after burning through $30 million.

And had you figured it out sooner, you would’ve done a better job with how you segmented your customers and/or how you adapted your product offering or commercial model, improving your chances of success.

What gets in the way of this insight?

Skill for sure. It is a skill and it is not equally distributed among founders and leaders.

But furthermore, the thing that gets in many people’s way is their mindset.

They wholeheartedly agree that others could fall prey to the customer problem blindspots above (smiling as they read), but they are sure that they are themselves immune to these blindspots, because they are so smart, so customer-centric, so persistent, such domain experts, so vision-driven, etc. etc.

They are certain they can read their customers’ minds, but alas they fail to read the tricks their own mind is playing on them.