Adisorn

[Adisorn @ad1sorn]

@elonmusk @theallinpod Only with the right political and monetary conditions supportive of AI+robotics investments.

[Adisorn @ad1sorn]

@chamath We’ve reached a point where significant % of business automation can be done without further model advancements.

Businesses have barely capitalized on existing intelligence. Lot of low hanging fruits.

[Adisorn @ad1sorn]

Spending a lot on tokens is not necessarily a good thing.

But spending too little on tokens will soon become an obvious sign of not using enough AI to stay competitive.

[Adisorn @ad1sorn]

@DavidSacks This will not just be a good idea but a necessity. With the upcoming compute crunch, squeezing the most intelligence out of smaller open models through post-training will be more common practice.

[Adisorn @ad1sorn]

@yoheinakajima 👋

[Adisorn @ad1sorn]

Everyone will compare Claude Opus 4.8 vs 4.7 on speed and cost.

The bigger story is autonomous execution, and whether your architecture can absorb it without a rebuild.

[Adisorn @ad1sorn]

6–12 months to payback is achievable.

But only if the first question you ask is: what does this problem cost us today?

Most teams skip that question.

That's why most teams are still waiting for results.

DM me if you want to work through the baseline calculation before your next AI project.

[Adisorn @ad1sorn]

The research is consistent across multiple 2026 studies:

AI automation delivers payback in 6–12 months, when the use case has a clear cost or revenue lever.

Most enterprise teams read "6–12 months" as a timeline.

I read it as a filter.

🧵

[Adisorn @ad1sorn]

You can't calculate ROI on a process you've never measured.

The manufacturing firms saving $1.2M annually on compliance reporting didn't get lucky.

They knew exactly what compliance reporting cost before they started. That number became the target.

[Adisorn @ad1sorn]

Enterprise AI transformation in 2026:

Product Owner: ChatGPT
Project Manager: Copilot
Engineers: vibe-coding tool
Testers: yet another assistant

Nobody: sharing context
Context: dying at every handoff

This is everyone moving fast in different directions.

[Adisorn @ad1sorn]

@stats_feed Checking your phone the moment you wake up

[Adisorn @ad1sorn]

@An0rexik It depends. The question is whether we'll be using it to work faster or to avoid thinking.

[Adisorn @ad1sorn]

@NoahKingJr Infrastructure

[Adisorn @ad1sorn]

@KaiXCreator They're taking turns. Whoever ships next leads for a few weeks.

The real question is which one fits your workflow 🙂

[Adisorn @ad1sorn]

@Vivek4real_ If intelligence becomes a utility, the competitive advantage shifts entirely.

Not who has the best model. Who built the best infrastructure around it.

[Adisorn @ad1sorn]

The fix: standardized workflows that capture expertise and enforce consistency.

One person's best output becomes everyone's baseline.

Not because everyone became an expert at prompting. Because the process does the work.

[Adisorn @ad1sorn]

Why does your team get inconsistent AI results?

Same tool. Same task. Completely different output quality.

It's not a skill problem. Here's what's actually happening.

[Adisorn @ad1sorn]

Sharing a prompt is not the same as sharing a workflow.

A prompt is an input. A workflow is a system.

One lives in a Notion doc. The other runs every time.