Lives by the "WHY" — questions the root cause until the philosophy is clear
Risk-first, capital preservation — trading run like an insurance business
Trading is a marathon — consistency across regimes, not daily heroics
Kapil Agarwal
Systematic Trader · Builder
"Control the unknowns. Diversify the positions. Engineer the failure modes out."
B.Tech in Computer Science · short stint in IT before going full-time on markets
Trading markets since 2017 — discretionary didn't work, went fully systematic
Migrating from no-code platforms to self-built systems
Built a multi-strategy, multi-broker execution stack — signal-based entries, trailing stops, crash recovery, event-sourced audit
Safety-first systems — missing SL, half-filled spreads, mid-day crashes — all handled. Instant Telegram alerts.
Hardened by live incidents — what breaks once never breaks the same way twice
In 12 months, AI went from assistant to coworker.
SEP 2025
Sonnet 4.5
30+ hours of autonomous coding
NOV 2025
Opus 4.5
Workplace automation
JAN 2026
Claude Cowork
Enterprise applications
FEB 2026
Opus 4.6 + Agent Teams
1M context · multi-agent
FEB 2026
COBOL Playbook
Legacy mainframe modernization
APR 2026
Opus 4.7
Self-verifying · vision · xhigh reasoning
Five major capability jumps in one year. We are not on a normal technology curve.
India's IT giants stopped denying it — they signed up.
The market priced the fall. Then the firms themselves confirmed the thesis —
partnering with the very company building the models that compress their billable hour.
Infosys
Signed · Front-runner
Strategic partnership with Anthropic. Claude & Claude Code embedded into Infosys Topaz.
Cognizant
Signed · Deepest adoption
Claude deployed to ~350,000 employees.
TCS
Announcing · India's largest
No formal deal yet. COO, April '26: TCS is "working significantly" with Anthropic. Formal announcement expected.
The chart was the market's verdict. These partnerships are the industry's confession.
November 2024 → May 2026
One company wrote $840 billion of new value.
The rest of the industry wrote a trillion-dollar funeral.
Anthropic valuationUS software market cap (indexed)
Feb 3, 2026
Claude Cowork launches
–$285B in a day
Feb 5, 2026
Opus 4.6 + Agent Teams
Salesforce –25% · SAP –18%
Feb 23, 2026
Claude Code COBOL Playbook
IBM –13% · $40B gone
Nov 2024
Series E
$60B
Feb 2026
Series G
$380B
May 2026
Talks underway
$900B
Nov '24Feb '25Jun '25Oct '25Feb '26May '26
The companies leaning away lost −$1T in market cap.
The one leaning in is up 15×.
The people building AI have been saying it for 3 years.
Five companies. One direction. Listen to the chorus.
JAN 2023
Tweet
"The hottest new programming language is English."
Andrej Karpathy — ex-OpenAI, ex-Tesla AI
FEB 2024
WGS Dubai
"Nobody has to program — the programming language is human. Everybody in the world is now a programmer."
Jensen Huang — CEO, NVIDIA
OCT 2024
Q3 Earnings
"More than 25% of new code at Google is now generated by AI — and reviewed by engineers."
Sundar Pichai — CEO, Google / Alphabet
JAN 2025
Joe Rogan
"In 2025, Meta will have an AI that is effectively a mid-level engineer. Most of our app code will be built by AI."
Mark Zuckerberg — CEO, Meta
MAR 2025
CFR Event
"In 3–6 months, AI writes 90% of code. In 12 months, essentially all of it."
Dario Amodei — CEO, Anthropic
Five different companies. Same direction.
LIVE
Pick one — 90 seconds. Build it in front of them.
1. Streamlit option-chain app — from one prompt to working code
2. Messy Excel → cleaned, pivoted, charted, summarized
3. P&L screenshot → trade analysis & what went wrong
Goal: the audible gasp.
This is the moment that converts skeptics.
While the giants fall, individuals are rising.
The leverage moved from organizations to individuals — the ones who learned to use the tools first.
The Solo Founder
Pieter Levels
Made
Photo AI — an app that makes professional portraits from your selfies. $1M a year, built and run by one person.
Also
Runs Nomad List & RemoteOK. ~$3M/year total. No team.
The Indie Studio
Marc Lou
Made
ShipFast — a starter kit that helps people launch a website in a day. Made $528K in 4 months.
Also
Sells 5+ small products. $1M+ in 2025. No team.
The Non-Coder
Kostiantyn Vlasenko
Made
Respiro — an app that spots when you're stressed and helps you calm down. Built in 6 weeks using Claude.
Also
Never wrote code before. Won Anthropic's coding contest.
The new asset class isn't capital — it's leverage through AI.
You don't get to opt out. You get to choose which side of the table.
Eight habits. Two columns. One transformation, line by line.
Coder yesterday
Vibe-Coder today
Types every line themselves
›
Tells the AI what to build
Memorizes syntax & shortcuts
›
Memorizes nothing — describes it
Stuck on one bug for hours
›
Asks Claude, fixes it in minutes
Googles error messages
›
Pastes errors into chat
Builds one feature a week
›
Ships a working prototype in a day
Paid for hours typed
›
Paid for problems solved
Code degree required
›
A good question is enough
Easy to replace
›
Hard to replace
Next Session · Paid Workshop
Your backtest said “buy.” Now what?
You downloaded data. You tested an idea. But between a backtest and real money, there's a machine you've never seen — orders, fills, stop-losses, rejections, race conditions. Next session, you build a piece of it.
Data
Session 1
→
Signal
Session 1
→
Order
Session 2
→
Fill
Session 2
→
Stop-Loss
Session 2
→
Trail
→
Squareoff
→
Recovery
→
Alerts
Next session: you build a simulator, then read a real system's trade log.
Session 2 — Two halves
Build it. Then read the real thing.
Hour 1 — Build a paper trading simulator
You describe a rule. Claude turns it into a simulator that runs on your NIFTY data. LIMIT orders, slippage, stop-losses, partial fills. Your backtest gets honest.
Hour 2 — Read a real production log
You get a day’s trade log from a system that trades real money. Parse it. Trace a trade. Spot the failure. Then see the live system on screen.
Sent 3 instead of 195
One segment counts in lots. Another in units. The API doesn't warn you. It just places the wrong quantity.
SL rejected at circuit limit
Exchange says max price is 180. Your SL was at 200. Rejected. Position now has no safety net.
Cancel and fill at the same time
You cancelled the order. Broker said OK. But it filled 30ms earlier. Now you own something you didn't want.
These are from a real bug log. 197 entries. You'll read three of them.
Next Session
From Signal to Order.
You build a simulator on your data. You read a real system's trade log. You see the live system on screen. No broker account needed.
One session. Hands-on.
~2 hours · you build + you investigate
₹2,500
Prerequisite: your NIFTY data from Session 1 (the free Parquet files).
