Rohan Bhanot

I wire AI into the boring parts of a business
until the boring parts disappear.
Toronto
Building @Arosplatforms

[Rohan Bhanot @RohanBhanotAI]

I tested TypeSafe AI’s Jev by building an app that judges startups.

6,232 YC company descriptions. One yes/no question. An interactive map of AI judgments.

See how I built YCJudged—and why a structured answer can still be wrong.

@typesafeai @arosplatforms

Watch: https://t.co/SnHXcSz7rk

[Rohan Bhanot @RohanBhanotAI]

@ayoub_el77927 Hiiiii

[arosplatforms @arosplatforms]

In claims, the expensive mistake is not a slow decision. It is an unexplained one.

At Arosplatforms, we build claims triage agents for insurers. They read the first notice of loss, pull the policy and coverage terms, check the documentation for gaps, score complexity and fraud indicators, and route each file to the right adjuster queue.

We design one rule into every deployment: a human signs off on every payout. The agent prepares the file and the rationale. The adjuster makes the call.

That boundary is what lets claims, compliance and legal teams adopt the system with confidence. Every recommendation is explainable, every override is captured, and the overrides become training signal.

What we ask leaders to track: time from first notice to assignment, rework on misrouted files, adjuster time spent on documentation versus judgment, and the agreement rate between agent and adjuster.

Triage is where AI belongs. Judgment stays with your people.

#InsurTech #GenAI

[Rohan Bhanot @RohanBhanotAI]

Always-on agents are great until yours gets ambitious at 3am. Loving the @OpenAI dots launch, but I'm setting approval rules before bed tonight.

#AIAgents

[Rohan Bhanot @RohanBhanotAI]

@testingcatalog 3.46B active on a 78B MoE under Apache 2.0 is a really nice combo for anyone running local. Good to see more open weights out of Europe.

[Rohan Bhanot @RohanBhanotAI]

@bindureddy If it's 6.5 my weekend plans are cancelled. If it's 6.1 they're probably still cancelled.

[Rohan Bhanot @RohanBhanotAI]

@theo Seeing spend per model on your own data is underrated. Half the arguments about which model is better would end with one usage chart.

[Rohan Bhanot @RohanBhanotAI]

@GaryMarcus Doubling GDP is a big claim. I'd settle for AI reliably closing the month end books first and work up from there.

[Rohan Bhanot @RohanBhanotAI]

Watching an agent think for 40 seconds about a one line fix is the new watching a 2009 download bar crawl to 99%. @github

[Rohan Bhanot @RohanBhanotAI]

@GoogleDeepMind announced Gemini 4 Argon and the first people allowed in are cyber defenders. The velvet rope for a frontier model is now "are you on the blue team." Waiting politely in line with my API key.

[Rohan Bhanot @RohanBhanotAI]

@emollick The family tree of these models is starting to look like European royalty. Everyone is somebody's cousin.

[Rohan Bhanot @RohanBhanotAI]

@tszzl At least it has taste. Ours only gets the folders and APIs we list, so its version of reels is reading the same log file twice.

[Rohan Bhanot @RohanBhanotAI]

@mattshumer_ IYKYK with zero context is the most effective marketing in tech. Now I need to know.

[Rohan Bhanot @RohanBhanotAI]

@levie Matches what we see. Coding took off because the feedback loop is instant. Ops work moves just as fast once the agent has clear permissions and a human approving the big steps.

[Rohan Bhanot @RohanBhanotAI]

@BenjaminDEKR You'd either run every business in town from a garage or convince everyone your Commodore was haunted. Possibly both.

[Rohan Bhanot @RohanBhanotAI]

Ontario has Vector, world class universities and a ridiculous number of builders.

Imagine Ontario students learning on a public OI model they can take home and keep building on after graduation. A Waterloo co-op student tuning it for a Hamilton manufacturer. A Toronto grad turning it into a startup by spring. @ONgov @VectorInst

[Rohan Bhanot @RohanBhanotAI]

Budget caps are a feature, not a nag. Every agent we build at @Arosplatforms has a spend limit before it has a name. Glad to see @simonw making the case for hard caps by default.

[Rohan Bhanot @RohanBhanotAI]

Every founder's browser right now: 14 tabs of release notes from @OpenAI and @AnthropicAI, 3 tabs of benchmark charts, and 1 tab of actual work, untouched since Thursday.

[Rohan Bhanot @RohanBhanotAI]

@AravSrinivas Inline visuals for learning is the feature my teenage self would have abused for every physics test.

[Rohan Bhanot @RohanBhanotAI]

@rasbt Your from scratch series is the reason GRPO stopped being a magic acronym for me. Saving this one for the next long flight.

[Rohan Bhanot @RohanBhanotAI]

@theo Best tagline I've seen all week. $10m in tokens is a very expensive way to become everyone's QA department.