Usual parents pay $20 for a plastic toy that costs $0,20 to create but smart parents buy 3D printer for $300 and create hundreds of toys every week for their children.
$300 appliance. $500/month saved. Potential business with $2,000+ net a month.
The manufacturer had stopped making spare parts years ago.
Buying a new machine was the only official solution.
Instead, someone printed one tiny gear.
The customer happily paid because replacing a small part is always cheaper than replacing the entire appliance.
That is why replacement parts are one of the best 3D printing businesses.
People don’t buy plastic.
They buy another five years of using something they already own.
you’re reading this on a device that could solve someone’s $800 problem before dinner
24 yo insurance analyst in chicago turned $400 into $5,100 in one week on a political event - no inside info, no advantage
spreadsheets by day, prediction markets at night
he didn't have better data. just had two tabs open at once
in spring he noticed same event priced 12 points apart across two separate prediction markets
he assumed gap would close in minutes. one market sat frozen for 90 minutes after a poll dropped
he nearly quit in month two - entered a gap, it went the wrong way first, down $180 before it reversed
what he found is called cross-market resolution drift: when new info drops, one market reprices fast, other lags behind
checkable now:
pull up two prediction markets on the same active event, compare implied odds
if spread is 8+ points and one side just had a volume spike, the other is probably still behind
her $400 -> $5,100 week was outlier. most runs were 6-9% when gap actually closed
prices were public whole time. nobody was checking both at once
you're not losing to better information - you're losing to the second tab you never opened
GOOGLE IS QUIETLY GIVING AWAY 15 AI TOOLS AND MOST PEOPLE ONLY KNOW GEMINI
The five worth your afternoon:
Pomelli — paste your website URL. It pulls your fonts, colors and tone, then builds a full social campaign that actually looks like your brand.
Stitch — one sentence in, production-ready UI and websites out. Still in beta and already replacing paid mockup tools.
Opal — n8n or Make, except Google's AI is baked in. Describe the workflow, it wires it up. No code.
Antigravity — an agentic code editor on Gemini 3 Pro. He asked for a flight lookup app, the agents planned it, built it, then generated the landing page for it too.
Mixboard — infinite canvas for images. Generate, remix, merge the parts you like, keep going until it matches what's in your head.
Brand, design, automation, code, visuals. Five tabs, zero dollars.
Most people are paying monthly for all five. Why?
A museum in Seoul put a neural network behind glass, and visitors line up to watch it think.
Under a steel sign 참여형 작품입니다, An Interactive Exhibit — sits a small touchscreen where a visitor draws a crooked 3 with one finger and hits Predict.
Then the wall lights up, and it doesn't show the answer it shows the process.
The 3 shatters into 784 pixels that fire into a lattice of thousands of glass-white cells floating on a 2-meter black screen. The lattice folds and stretches, pushing the scribble through layer after layer, each one crushing 784 numbers down until only 10 remain 1 per digit, and the biggest one wins.
The real computation takes 4 milliseconds, but the museum slows it to 20 seconds so a human eye can follow.
This is MNIST 70,000 handwritten digits collected in the 1990s from US Census workers and high schoolers, the dataset every ML student meets first. Most people run it in a terminal and see 1 line: accuracy 98%.
This museum built it a body.
Kids drew sloppy 7s and laughed when it guessed 1, and that's exactly the point the machine stops being magic and becomes a stack of small dumb steps that add up to reading.
Every model you talk to Claude, GPT, the thing drafting your emails is this same lattice with 1,000,000x more cells.
You just can't watch those think yet.
CHATGPT WORK COULD OUTSHINE CLAUDE COWORK
Not because it has the “best model”
but because OpenAI has brought everything together in a single app
Now ChatGPT can do more than just answer questions.
It can:
▸ read documents
▸ gather context from Notion, Slack, and Drive
▸ analyze spreadsheets
▸ create presentations
▸ generate PDFs and XLSX files
▸ turn files into finished work products
Previously, there was a huge gap between regular ChatGPT and Codex.
Now, Work bridges that gap.
→ Chat — for thinking and writing
→ Work — for creating work products
→ Codex — for working with code
But there’s a catch:
Work and Codex share the same quota.
In other words, the new mode doesn’t give you more resources.
It just gives you another way to use them.
I’ve figured out what’s really changed in the new ChatGPT and why choosing a mode is now more important than choosing a model ↓