Prashant Chandel

Bias for Action | AI/ML | UT Austin | IIT Delhi

Daily strategy breakdowns built with verified data and agentic systems.

[Prashant Chandel @PC0697]

📌 Trends don't make business history. It's made by decisions.

This account studies those decisions.

Forget the fluff. We provide compressed, judgment-focused breakdowns of the choices that defined companies.

It’s like a mini-MBA in your feed, distilled by advanced AI agents that plan, critique, and rewrite until only the truth remains.

Follow to learn how to think, not just what to think. 🧠

👇 Start reading the threads below.

[Prashant Chandel @PC0697]

Both things can be true. The outcome depends on whether per-token cost compression (driven by on-device AI and hyperscaler competition) outpaces token consumption growth (driven by agentic use cases). We will know by Q4 2026 when enterprise budgets get renewed or slashed.

[Prashant Chandel @PC0697]

Microsoft cancelled most internal AI coding licenses after 6 months. One Nvidia exec explained: "the cost of compute was far beyond the cost of the employees." This is Microsoft - the company that owns 49% of OpenAI. Even they could not make the math work.

[Prashant Chandel @PC0697]

Agentic AI is still early. As models improve reasoning efficiency, task completion could require fewer tokens. A task burning 10,000 tokens today might only need 2,000 tokens in 2027. Better models could solve the consumption problem before it becomes systemic.

[Prashant Chandel @PC0697]

The big question: do token costs compress faster than consumption grows, or is the enterprise AI market about to hit an ROI wall? The Microsoft and Uber cancellations suggest we are closer to the wall than the industry wants to admit. Watch Q4 2026 budget announcements.

Full piece: https://t.co/MPDpdLsVvO #AI #Strategy #Tech

[Prashant Chandel @PC0697]

SaaS companies are shifting to consumption-based pricing - charging per AI task instead of per user seat. This is not margin expansion. It is cost-shifting. They are passing unpredictable token costs to customers to avoid being crushed by usage spikes.

[Prashant Chandel @PC0697]

Personal opinion: if you are getting one of the IIT's join it and put the effort there instead of dealing with uncertainty for the next year.

At different points in life you will feel you made the wrong or the right choice.

10 years down the line either way wont matter if you continue to work hard.

PS: I took a drop

[Prashant Chandel @PC0697]

AI just created a new attack vector called slopsquatting.

AI hallucinates a package name.

The attacker registers it.

The developer installs it on the AI's recommendation. Malware ships.

1 in 5 AI-generated code suggestions references a package that does not exist.

[Prashant Chandel @PC0697]

@mert https://t.co/MIJR7C6HOK

[Prashant Chandel @PC0697]

Long Form Content >> Most tweets

No contrarian takes just to get more engagement.

[Prashant Chandel @PC0697]

@NaivaidyaY66600 Fail fast, fail forward

[Prashant Chandel @PC0697]

@_fernando_rosas It has been post trained on the responses that humans would like so its goal should be to add value to the person inputting the tokens

[Prashant Chandel @PC0697]

@paulg @t_blom New Companies will eventually become more efficiently run organizations?

[Prashant Chandel @PC0697]

The Solow Paradox redux: In the 1970s, massive IT investment failed to boost productivity because companies automated the wrong work.

Now: 95% of AI pilots fail for the same reason. They make broken processes faster instead of fixing them first.

[Prashant Chandel @PC0697]

Long Form Content >> Most tweets

No contrarian takes just to get more engagement.

[Prashant Chandel @PC0697]

@YashHustle_22 Tyler and Cameroon Winklevoss would have been much happier going to school in 2026.