AI agents can shop for you now. They can find a product, compare prices and check out in seconds. But an agent that can spend your money with no limits is a risk nobody should take. One bad search, one sketchy merchant, one purchase on the wrong day, and the money is already gone.
This is why we built π π’π§πππ₯ for the @PayPal AI Hackathon.
FinPal is an AI shopping agent that works inside your spending rules and never pays until you approve.
Here's how it works:
β You type a list, snap a product photo or just say what you need
β The agent searches live offers and compares prices against the market
β Your rules decide what can be bought: per-purchase caps, daily and monthly limits, and the days and hours spending is allowed
β If a merchant looks off, it doesn't guess. It stops and asks you to review
β Nothing is paid until you approve it through PayPal checkout
β You can also send money to people, and the same rules apply
Every purchase keeps its evidence: the price check, which rules passed, and a plain-words summary. Activity keeps a record that can't be edited, so nothing happens out of sight.
The AI explains every decision. It never makes them. Your rules decide, and you approve.
Built with PayPal Orders v2 and Payouts, @channel3_ai , Next.js, Supabase and Groq, all running on PayPal Sandbox.
Been heads down on this with @iamtechhunter@ikeh_chukw38247@CeeHenrie for the past couple of weeks, and we're excited to finally share it π
Watch the demo π
https://t.co/Au4FA2X7PY
Devpost: https://t.co/gS8XSjJYBb
Try it: https://t.co/K5htm5PKz3
I built ChogChoir for #Chogathon: every Chog has its own voice and song, made from its real traits.
Hum a tune and it sings it back, pair two for a duet, and sing your Chog into a shared on-chain choir on Monad. Encores burn $CHOG.
Try it: https://t.co/yKBEC7b91p
Day 1: you start work.
Day 30: you finally get paid.
What if pay landed every second instead?
meet Payday π
Tell Claude who to pay. Approve once in @PayBox. Pay streams on Base, by the second.
You ask. We handle payday
Our new age - consumer finance
Day 32 of #GeekStreak60
βLargest BSTβ had me stuck for a bit π
Tried to find BSTs manuallyβ¦ didnβt work.
Switched approach β let each subtree decide if itβs a BST.
That changed everything.
#npci#coding#algorithms#GeekStreak60
Day 25.
Minimum K consecutive bit flips.
Instead of flipping bits every time, track flip state with a sliding window.
If current bit (after flips) is 0 β start a flip.
Greedy + difference array trick.Time: O(n)
#geekstreak60#npci#DSA
Day 23.
Subarrays where the first element is the minimum.
Key idea:
Find the next smaller element using a monotonic stack.
If next smaller index is j,
valid subarrays starting at i = j - i.
Time: O(n)Stack patterns everywhere.
#geekstreak60#npci#DSA
Day 22.
Largest number in one swap.
Track last position of digits 0β9.
From left β find a bigger digit later β swap once.
Greedy + indexing trick.O(n).
#geekstreak60#npci#DSA
Open Source Automation Project
One student built an automation workflow used in an open-source project.
The workflow supports contributor coordination in GitPadi.
Real projects create stronger learning.
https://t.co/bKFVJRIiat
Day 19.
Smallest window containing all characters.
Sliding window again.Count chars in p.
Expand right.
Shrink left when all chars are matched.
Track the smallest window.
Time: O(n)Patterns keep repeating in DSA.
#geekstreak60#DSA#CodingJourney#npci
Built Maverick β an autonomous multi-agent wallet ecosystem on Solana where AI agents create their own wallets, sign transactions, trade, swap, lend, and talk back to you.
Every operation is a real on-chain transaction on devnet. No mocks.
Thread on what it does
Security matters even in a prototype:
- Each agent has an isolated keypair
- Vault keypair is separate only it can sign outbound vault txs
- Append-only audit trail with tx signatures
- Block height expiry handling verifies tx landed on-chain before reporting failure