I Monitor 78 AI Channels So You Don't Have To. Here's What Actually Matters.

posted Originally published at medium.com 2 min read

A guy made a cancer vaccine for his dog using ChatGPT. A realtor lost $24K in commissions to AI. And Anthropic quietly doubled their limits.

I spend 30 minutes every morning reading AI news from 78 Telegram channels. Most of it is noise. "This changes everything." "The future is here." You know the type.

But every week, 3 or 4 things actually matter. Not because they are cool. Because they change how real people make money or lose it.

Here is what stood out this week.

The $24,000 realtor story

A guy in Florida sold his house without a realtor. Used ChatGPT for everything. Market analysis, repair priorities, listing description, marketing flyers. Took him 5 days. Saved $24,000 in commission.

He still needed a lawyer for the closing. But the $24K that normally goes to an agent? Gone.

This is the pattern I keep seeing. AI does not replace the expert. It replaces the $200/hour expert who was doing $40/hour work. The strategic thinking stays human. The busywork disappears.

The dog cancer vaccine

An IT guy in Australia found out his rescue dog Rosie had cancer. No veterinary oncologist would help with experimental treatment. So he opened ChatGPT and AlphaFold, taught himself mRNA vaccine design with zero biology background, and built a custom vaccine for his dog.

Three months later the tumor shrank by half.

I am not saying everyone should make vaccines at home. But this is what happens when information stops being locked behind credentials. The tools are the same ones Moderna uses. The difference is a $0 budget and a guy who refused to accept "nothing we can do."

What Anthropic did that nobody noticed

First, they opened the full 1 million token context window on Claude. That is roughly 600 pages in a single request. No extra charge. For context, most people are still working with 8K or 32K windows and hitting limits every 10 minutes.

Why does this matter? Because the biggest limitation of AI for business was always "it forgets what I told it 5 minutes ago." With 1M tokens, you can load an entire quarter of customer emails, contracts, and reports into one conversation. The AI sees everything at once. No more feeding it information piece by piece.

Second, they quietly doubled usage limits during off-peak hours. Nights and weekends, you get twice the capacity. If you run automations (I run 7 every morning before I wake up), this is free money.

The noise you can ignore

"This new model scores 3% higher on benchmarks." Nobody cares. Your customers do not buy benchmarks. They buy results.

"AI will replace all jobs by 2027." It will not. It will replace the boring parts of some jobs. The parts nobody wanted to do anyway. I replaced a $4,000/month employee with AI. The AI does the research and sends the emails. A human still closes the deal.

The real question is not "will AI take my job?" It is "am I spending my expensive human hours on $40 tasks?"

If yes, you are the next realtor losing $24K to a guy with a laptop and ChatGPT.


Dmytro Negodiuk is a Fractional AI Officer based in NYC, helping startups and mid-size businesses build AI systems that work. He runs 5 businesses with zero employees using custom AI systems.

Website: negodiuk.ai | LinkedIn: linkedin.com/in/dima-negodiuk

1 Comment

0 votes

More Posts

5 Things This Playwright SQL Fixture Does So You Don't Have To

vitalicset - Apr 13

I’m a Senior Dev and I’ve Forgotten How to Think Without a Prompt

Karol Modelskiverified - Mar 19

AI Agents Don't Have Identities. That's Everyone's Problem.

Tom Smithverified - Mar 13

Everyone says DeepSeek is cheaper, but I got tired of guessing the exact math. So I built a calculat

abarth23 - Apr 27

Your AI Agent Skills Have a Version Control Problem

snapsynapseverified - Apr 22
chevron_left

Commenters (This Week)

3 comments
2 comments
1 comment

Contribute meaningful comments to climb the leaderboard and earn badges!