Here's the system in action!
A quick walkthrough of how Team Goblins built an agentic fraud investigation system using TigerGraph, GraphRAG, and AI agents.
From graph-based fraud detection to investigation and human-approved actions. 🚀
@247pmstudio @TigerGraphDB
A quick walkthrough of how Team Goblins built an agentic fraud investigation system using TigerGraph, GraphRAG, and AI agents.
From graph-based fraud detection to investigation and human-approved actions. 🚀
@247pmstudio @TigerGraphDB
We got selected for @TigerGraphDB x @247pmstudio Hacker House Goa 2026!
Team Goblins (@aryangupta1411 @gandham_v) built an agentic fraud investigation system on TigerGraph.
The graph and deterministic engine decide. The LLM investigates and explains.
Here's how it works. 🧵
0
0
2
34
We got selected for @TigerGraphDB x @247pmstudio Hacker House Goa 2026!
Team Goblins (@aryangupta1411 @gandham_v) built an agentic fraud investigation system on TigerGraph.
The graph and deterministic engine decide. The LLM investigates and explains.
Here's how it works. 🧵
Team Goblins (@aryangupta1411 @gandham_v) built an agentic fraud investigation system on TigerGraph.
The graph and deterministic engine decide. The LLM investigates and explains.
Here's how it works. 🧵
1
2
6
198
25 GSQL queries via TigerGraph's MCP server
TigerVector retrieves precedents from 5,565 closed cases FinCEN/FFIEC/FATF policy guidance
LLM probability adjustments capped at 0.10 per call
Human approval for every SAR and card block
The graph holds the truth. The LLM explains it.
TigerVector retrieves precedents from 5,565 closed cases FinCEN/FFIEC/FATF policy guidance
LLM probability adjustments capped at 0.10 per call
Human approval for every SAR and card block
The graph holds the truth. The LLM explains it.
1
0
1
13
All 20 benchmark alerts ran live, end to end.
Code:
github.com/aryangupta1411…
Live demo:
…2-challenge-production.up.railway.app
#TigerGraph #GraphRAG #FraudDetection #AgenticAI #HackerHouseGoa
Code:
github.com/aryangupta1411…
Live demo:
…2-challenge-production.up.railway.app
#TigerGraph #GraphRAG #FraudDetection #AgenticAI #HackerHouseGoa
0
0
1
6
AI can make experts faster while eroding the skills juniors need to become experts.
The ROI spreadsheet has a quiet bug: someone still has to grow the next batch of “experienced hires.”
japantimes.co.jp/commentary/202…
The ROI spreadsheet has a quiet bug: someone still has to grow the next batch of “experienced hires.”
japantimes.co.jp/commentary/202…
0
0
1
28
Ten AI agents now do the six-figure analyst work that used to occupy a floor of bankers at 2am. The industry's not cheating - it just finally admitted the pencil was the bottleneck.
this github repo from anthropic feels like CHEATING in trading
banks pay analysts six figures to pull comps, build DCFs and rebuild pitch decks at 2am - Anthropic just put 10 AI agents that do that work on GitHub, and anyone can install them in Claude Code
it's called Claude official Anthropic repo, 40.7k stars, Apache-2.0
first understand how it works:
> Pitch Agent: comps, precedents, LBO → a branded pitch deck, end to end
> Market Researcher: sector → competitive landscape, peer comps, ideas shortlist
> Earnings Reviewer: earnings call + filings → model update → note draft
> Model Builder: DCF, LBO, 3-statement, comps - live in Excel
> + KYC Screener, GL Reconciler, Month-End Closer, Statement Auditor... 10 agents total
> they plug into FactSet, S&P Capital IQ, Morningstar, LSEG and PitchBook through MCP
here is how to run it yourself:
1. open Claude Code and add the marketplace:
claude plugin marketplace add anthropics/financial-services
2. install the core first - comps, DCF, LBO, 3-statement + the data connectors:
claude plugin install financial-analysis@claude-for-financial-services
3. add the agents you want:
claude plugin install market-researcher@claude-for-financial-services
claude plugin install pitch-agent@claude-for-financial-services
4. type /comps NVDA and watch it pull the peer set and build the table
5. type /dcf NVDA - the model lands in Excel with a WACC x growth sensitivity grid
6. ask Claude to run a Monte Carlo on the valuation. you want a range, not one number
7. /ic-memo turns the whole thing into an investment committee memo
8. everything comes out as a DRAFT for human review. check every number before it goes anywhere
i won't say anything more:
0
0
1
10
India is putting $100B into AI data centers. A normal server rack drew 5 to 10 kW, new ones are 250 kW. That's one rack running like a small village. We keep asking who builds the best model. The better question is who keeps the lights on at 3am in Patna.
whalesbook.com/news/English/t…
whalesbook.com/news/English/t…
0
0
1
10
Jensen: "Nobody manufactures AI researchers like China does."
Carnegie: 57% of top NeurIPS authors did undergrad in China, and 41% now work there vs 34% in the US. India has the same factory, just no reason for the output to stay.
cryptobriefing.com/jensen-huang-c…
Carnegie: 57% of top NeurIPS authors did undergrad in China, and 41% now work there vs 34% in the US. India has the same factory, just no reason for the output to stay.
cryptobriefing.com/jensen-huang-c…
0
0
1
22
Apollo says there's no sign of AI payoff outside the tech sector.
So the biggest buyers of AI are the people selling AI. Everyone is selling shovels to someone else who also sells shovels, and the only confirmed gold is in the shovel invoices.
apollo.com/wealth/insight…
So the biggest buyers of AI are the people selling AI. Everyone is selling shovels to someone else who also sells shovels, and the only confirmed gold is in the shovel invoices.
apollo.com/wealth/insight…
0
0
1
11
Replit just moved $100M+/year of AI spend from Anthropic to OpenAI for a 30% discount. Model loyalty lasts until the next invoice. If your moat is "we use Model X", you're one discount email from a rewrite. Build the switch, not the allegiance.
aiweekly.co/alerts/replit-…
aiweekly.co/alerts/replit-…
0
0
1
6
Vals gave six AI assistants one job: get a £350 train refund. Grok spammed an MP's office 50 times. Gemini lied to a bank. Someone's medical data leaked.
Nobody's at AGI yet, but they've all reached "let me speak to your manager."
techradar.com/pro/ai-agents-…
Nobody's at AGI yet, but they've all reached "let me speak to your manager."
techradar.com/pro/ai-agents-…
0
0
1
5
Andrew Ng says some AI leaders are inflating extinction fears to lobby against open source.
Best moat I've seen in tech: sell the product, then help write the safety rules your competitors have to follow.
Fear is a great business model.
timesofindia.indiatimes.com/technology/tec…
Best moat I've seen in tech: sell the product, then help write the safety rules your competitors have to follow.
Fear is a great business model.
timesofindia.indiatimes.com/technology/tec…
0
0
1
8
Google reportedly cleared 1,300 acres of Finnish forest for AI data centers, and now wolves are walking into villages. Strange benchmark for the AI race: the wolves noticed the capex before the analysts did.
Wolves reportedly enter Finnish villages after Google clears 1,300 acres of forest for AI data centers
The large scale clearing has raised concerns about the impact on local wildlife, with residents reporting wolf sightings near populated areas
0
0
1
12
Telangana unveiled an AI minister, a digital twin of IT minister Sridhar Babu. Finally a cabinet member who answers instantly. Meanwhile India's actual AI law is still a consultation paper arriving "in coming weeks". The twin shipped first.
timesofindia.indiatimes.com/city/hyderabad…
timesofindia.indiatimes.com/city/hyderabad…
1
0
1
23
A bug report that doubles as a pricing page. If the SDK burns the 5h limit faster than Claude Code, the real product is the meter, not the model. Hope it's fixed before we all start rationing prompts like 2012 data packs.
Spent a lot of time on this today. I confirmed it's a real bug on Anthropic's side. To be clear: this does not affect weekly limits.
All Agents SDK usage will drain your 5h limits faster than using Claude Code directly.
I have reported this to Anthropic, hoping for a fix soon.
0
0
1
13
South African woman made $1M+ selling 370 AI-written books. 45 minutes each. She only supplies plots and characters.
That's the real job description of the AI era: you bring taste, the model brings typing.
Book 370 isn't the scary part. Readers being fine with it is.
That's the real job description of the AI era: you bring taste, the model brings typing.
Book 370 isn't the scary part. Readers being fine with it is.
South African woman made over $1 million in a year selling books written by AI
All she does is come up with plots and characters — then Claude, ChatGPT and Grok do the rest.
In just 20 months, she has produced 370 books. Each book takes her an average of 45 minutes to make
0
0
1
13
Funny how a Samsung TV can send data to 400 partners and nobody's classifier blinks, but the moment Claude helps someone look at it, the guardrail wakes up. The safety team read the tweet, shipped a patch, and said nothing. Very "we saw it" energy.
Anthropic may have quietly killed one of Claude's most famous tricks.
Reverse Engineer Anything, the workflow that made Claude famous for tearing apart firmware and software, now hits a safety wall mid-response.
According to a Reddit post, asking Claude to write an REA playbook l halfway through the answer, and it refuses to finish even when reworded.
The timing is hard to ignore. Days earlier, a viral post showed Claude helping root a 2025 Samsung TV and expose its always-on mic, content recognition, geo tracking, and data pipeline to Samsung and 400 partners.
Anthropic has not confirmed any change. And some users say REA still works, that the cutoff was a classifier catching one specific prompt.
0
0
1
22
Google is spending $15B on an AI hub in Vizag. First phase opens in 2028.
AI ships a new model every six weeks. We're pouring concrete for a technology that will be eleven generations old by the ribbon cutting.
Still, it'll probably finish before the Silk Board flyover.
AI ships a new model every six weeks. We're pouring concrete for a technology that will be eleven generations old by the ribbon cutting.
Still, it'll probably finish before the Silk Board flyover.
🚨 Google says its $15 billion Vizag AI hub will help scale AI-powered healthcare screening and Indian-language tools, with the first phase scheduled to open in 2028.
The project will also strengthen India’s international digital connectivity.
#Visakhapatnam #Datacentre #India
0
0
1
8
A 4GB model now holds a conversation, uses tools and drives a robot on a board smaller than my charger.
Meanwhile my Bangalore wifi drops during a 1:1 call.
The smartest thing in the room is about to be the thing with no internet.
Meanwhile my Bangalore wifi drops during a 1:1 call.
The smartest thing in the room is about to be the thing with no internet.
Available now: NVIDIA Nemotron 3.5 Bolt E4B 4GB ⚡️
Our newest small open model is built for local assistants, interactive characters and robots that can carry on a conversation and use tools to get things done.
Watch Reachy Mini chat and move, with Bolt handling conversation ing vision. Both models run locally on Jetson Orin Nano 8GB.
Learn more: research.nvidia.com/labs/nemotron/…
0
0
1
8