Darshan Tank

AI + Data guy @SentientAGI | prev-@GnaniAi, @deloitte

[Darshan Tank @TankDarshan7]

@omarsar0 Thank you very much for sharing my work!

check it out: https://t.co/J6dPhVqsBK

[elvis @omarsar0]

Are agent skills always worth using?

The answer is no?

This paper provides some important insights to understand this more.

(bookmark it)

Paper summary:

Adding procedural skills to an agent is usually scored by average task success. That number nets gains against damage and hides half of what happened.

Setup: nearly 6,000 paired runs across two office automation benchmarks and three model harness stacks, comparing the same agent with and without skills.

A regression is a task the agent solved without skills and then failed once skills were added. Regressions are large enough that the best performing skills separate themselves mainly through fewer regressions. Larger gains contribute much less.

Three mechanisms drive it.

Skill description osmosis, where a skill changes agent behavior just by sitting in context even when it is never invoked. Grounding displacement, where a prescribed procedure overrides how the agent reads its inputs. Verification displacement, where the procedure suppresses checks the agent would otherwise run on its own output.

Trace analysis surfaces that procedural guidance is the stage least often responsible for failure, while grounding and verification dominate the errors that remain. Existing skills are almost entirely procedure.

Paper: https://t.co/6dNPyN6ali

Learn to build effective AI agents in our academy: https://t.co/1e8RZKs4uX

[Darshan Tank @TankDarshan7]

We got gta 6 before gta 6

[Darshan Tank @TankDarshan7]

@thsottiaux True! its clearly visible in your profile xD

[Darshan Tank @TankDarshan7]

@akokoi1 isn't he looks like Dinesh from silicon valley in real😂

[Darshan Tank @TankDarshan7]

Voice mode on @claudeai is fucked compared to @ChatGPT !!!!!

Dude who reads out intermediate reasoning!

[Darshan Tank @TankDarshan7]

Every piece of work you do has hidden rewards that you get to see by the time!

[Darshan Tank @TankDarshan7]

I don’t understand why FDE is so hyped up!
It’s basically client facing role where you help your client integrate and use AI!
Same shit we used to do at gnani ai back in 2023-2024!

[Sentient Ecosystem @SentientEco]

“CryptoAnalystBench: Failures in Multi-Tool Long-Form LLM Analysis” has been accepted as a poster at KDD 2026 (@kdd_news) — one of the leading conferences for AI, machine learning, and data mining.

At KDD? Meet with Sentient researcher Darshan Tank (@TankDarshan7) to learn why LLM agents struggle with multi-tool, long-form analysis in crypto research.

[Darshan Tank @TankDarshan7]

@archiexzzz Why don’t you share it on https://t.co/Ilwzm5bpnJ

[Darshan Tank @TankDarshan7]

@jeffr_yyy @deepeval I've seen demand for voice ai evals recently, excited!

[Darshan Tank @TankDarshan7]

@GeminiApp is boon when you are in non English speaking country!

[Darshan Tank @TankDarshan7]

Gemini for day to day search and specifically on Google Map is the best thing!

[Sentient @SentientAGI]

Agents can fetch data and nail the facts, but combining it all into real research is where they start to break.

Sentient researchers Darshan Tank (@TankDarshan7) and Sidhant Rahi tackle this in "CryptoAnalystBench: Failures in Multi-Tool Long-Form LLM Analysis", accepted into KDD 2026 (@kdd_news) in Jeju, South Korea ↓

[Darshan Tank @TankDarshan7]

@prathoshap I created skill that help you create poster from your research paper after asking you few questions around your preferences.

Happy to share it via skillpatch

https://t.co/AJq4rOghpe
https://t.co/1ekQh1i424

[Darshan Tank @TankDarshan7]

@PackBropagated Yeah, Opus 5 for diagrams is better than before.

[Darshan Tank @TankDarshan7]

One thing I observed, Opus 5 is good at creating diagrams

[Darshan Tank @TankDarshan7]

Our mind understands structure information better!

[Sentient @SentientAGI]

Everyone assumes more skills = better agents.

But Sentient researchers Darshan Tank (@TankDarshan7) and Baran Nama found that adding skills to an agent can break tasks it used to solve.

In their paper "The Regression Tax", this hidden cost wiped out 59% of the gains that skills produced across nearly 6,000 runs ↓

[Darshan Tank @TankDarshan7]

@Kseniase_ Thank you for sharing my work :)

[Darshan Tank @TankDarshan7]

It takes less than what we have in our mind!