Dain (@Dain0x) header
Dain (@Dain0x) profile picture

24 | building real things with AI, not threads about it

Joined September 2026
Photo by Dain (@Dain0x)Photo by Dain (@Dain0x)Photo by Dain (@Dain0x)Photo by Dain (@Dain0x)Photo by Dain (@Dain0x)Photo by Dain (@Dain0x)Photo by Dain (@Dain0x)Photo by Dain (@Dain0x)Photo by Dain (@Dain0x)Photo by Dain (@Dain0x)
Dain retweeted
I BUILT A TERMINAL THAT TRACES WHO FUNDED 20 WALLETS. The first address funded 8 of them. Two addresses funded 14. By the end of the scan, all 20 led back to just 4 funding sources. Here’s what the terminal does: → INGEST loads 20 wallet addresses and their transfer records. → TRACE follows each wallet’s funding link back to the sending address. → GROUP puts wallets with the same funding source on the same branch. → MAP shows how much of the wallet set each source funded. The final breakdown: Source 01 → 8 wallets → 40% Source 02 → 6 wallets → 30% Source 03 → 4 wallets → 20% Source 04 → 2 wallets → 10% This run uses simulated data. A shared funding source can be an exchange, a distributor, or a single operator. The graph gives you the addresses to investigate next. And that’s where I’d run the second scan: who funded those four addresses? WALLET TRACE / DAIN
2
2
18
544
I BUILT A TERMINAL THAT TRACES WHO FUNDED 20 WALLETS. The first address funded 8 of them. Two addresses funded 14. By the end of the scan, all 20 led back to just 4 funding sources. Here’s what the terminal does: → INGEST loads 20 wallet addresses and their transfer records. → TRACE follows each wallet’s funding link back to the sending address. → GROUP puts wallets with the same funding source on the same branch. → MAP shows how much of the wallet set each source funded. The final breakdown: Source 01 → 8 wallets → 40% Source 02 → 6 wallets → 30% Source 03 → 4 wallets → 20% Source 04 → 2 wallets → 10% This run uses simulated data. A shared funding source can be an exchange, a distributor, or a single operator. The graph gives you the addresses to investigate next. And that’s where I’d run the second scan: who funded those four addresses? WALLET TRACE / DAIN
2
2
18
544
Dain retweeted
TWO SKELETONS. WATCH THEIR FEET. ↑ GPT‑6.1 Sol — I built this version with Codex and Three.js, from the first stumble to a jump over a hurdle. ↓ Fable 5.2 — the original reference clip. Watch what happens when each foot hits the floor. Does it hold its ground? Does the body shift over it? That’s where the difference shows. Different setups. Which one moves more like a real person?
Claude Fable 5.2 is the most insane AI model in history... 99% of people will never touch it or burn money on it: "Too expensive", "Not for me" Use it right, and it becomes your second BRAIN, for a fraction of the price In this article, I show you how:
2
1
20
1,364
TWO SKELETONS. WATCH THEIR FEET. ↑ GPT‑6.1 Sol — I built this version with Codex and Three.js, from the first stumble to a jump over a hurdle. ↓ Fable 5.2 — the original reference clip. Watch what happens when each foot hits the floor. Does it hold its ground? Does the body shift over it? That’s where the difference shows. Different setups. Which one moves more like a real person?
Claude Fable 5.2 is the most insane AI model in history... 99% of people will never touch it or burn money on it: "Too expensive", "Not for me" Use it right, and it becomes your second BRAIN, for a fraction of the price In this article, I show you how:
2
1
20
1,364
Dain retweeted
3 AI MODELS. ONE EARTH. ONE PROMPT THE DIFFERENCE IS HARD TO IGNORE. The creator says all three received the same prompt. → Top: Gemini 4 Hydrogen → Middle: GPT-6.2 Sol → Bottom: Claude Fable 5.2 Watch the globe, then watch the details: the lighting, the camera, the way the interface holds together as the view changes. For me, Fable wins this visual comparison. The depth and lighting make the whole scene feel more convincing. One example doesn’t settle which model is best. But it does make a good blind test: Hide the names. Which one would you pick?
Claude Fable 5.2 is the most insane AI model in history... 99% of people will never touch it or burn money on it: "Too expensive", "Not for me" Use it right, and it becomes your second BRAIN, for a fraction of the price In this article, I show you how:
3
1
33
4,404
3 AI MODELS. ONE EARTH. ONE PROMPT THE DIFFERENCE IS HARD TO IGNORE. The creator says all three received the same prompt. → Top: Gemini 4 Hydrogen → Middle: GPT-6.2 Sol → Bottom: Claude Fable 5.2 Watch the globe, then watch the details: the lighting, the camera, the way the interface holds together as the view changes. For me, Fable wins this visual comparison. The depth and lighting make the whole scene feel more convincing. One example doesn’t settle which model is best. But it does make a good blind test: Hide the names. Which one would you pick?
Claude Fable 5.2 is the most insane AI model in history... 99% of people will never touch it or burn money on it: "Too expensive", "Not for me" Use it right, and it becomes your second BRAIN, for a fraction of the price In this article, I show you how:
3
1
33
4,404
Dain retweeted
THE SKY AND THE OCEAN SWITCH PLACES. YOU NEVER STOP MOVING. 21 luminous arches stretch across dark water. Their reflections form a tunnel that seems to go on forever. You fly toward the opening. Then the camera rolls 180°. The water moves overhead. The arches turn upside down. What felt like a clear path becomes a world with no obvious up or down. No cut. No scene change. Just one continuous shift in perspective. Watch the waterline through the turn. It tells the whole story. INVERTED / DAIN Built with code. Designed to bend your sense of space.
Grok Bot just got a serious brain upgrade With Claude Opus 5.5 entering the stack, one Bot could run an entire AI team that researches, writes code, uses tools, verifies results, and delivers finished work Here's how to build it in 10 steps:
1
17
1,748
THE SKY AND THE OCEAN SWITCH PLACES. YOU NEVER STOP MOVING. 21 luminous arches stretch across dark water. Their reflections form a tunnel that seems to go on forever. You fly toward the opening. Then the camera rolls 180°. The water moves overhead. The arches turn upside down. What felt like a clear path becomes a world with no obvious up or down. No cut. No scene change. Just one continuous shift in perspective. Watch the waterline through the turn. It tells the whole story. INVERTED / DAIN Built with code. Designed to bend your sense of space.
Grok Bot just got a serious brain upgrade With Claude Opus 5.5 entering the stack, one Bot could run an entire AI team that researches, writes code, uses tools, verifies results, and delivers finished work Here's how to build it in 10 steps:
1
17
1,748
Dain retweeted
WHAT THE F*CK AM I FLYING THROUGH? 120,000 PARTICLES. 5 SPIRAL ARMS. 12 SECONDS INSIDE THE STORM. At first, you can only see what’s rushing past you. Gold dust. Blue sparks. Streams of light disappearing beyond the frame. Then the camera turns. Those separate streams belong to something much bigger. Built with code. Every particle rendered. An imagined universe you can fly through. Watch the gaps between the spirals. That’s where the scale starts to hit. STARFALL / DAIN
An AI employee audited my AI bill and roasted me harder than my followers ever have. I was burning $4,700 a year on tools I barely opened. My dumbest habit was asking ChatGPT what Grok thought, and paying for both. This is the system that fixed it 👇
1
1
15
9,265
WHAT THE F*CK AM I FLYING THROUGH? 120,000 PARTICLES. 5 SPIRAL ARMS. 12 SECONDS INSIDE THE STORM. At first, you can only see what’s rushing past you. Gold dust. Blue sparks. Streams of light disappearing beyond the frame. Then the camera turns. Those separate streams belong to something much bigger. Built with code. Every particle rendered. An imagined universe you can fly through. Watch the gaps between the spirals. That’s where the scale starts to hit. STARFALL / DAIN
An AI employee audited my AI bill and roasted me harder than my followers ever have. I was burning $4,700 a year on tools I barely opened. My dumbest habit was asking ChatGPT what Grok thought, and paying for both. This is the system that fixed it 👇
1
1
15
9,265
Dain retweeted
1 signal. 12 seconds. An entire city brought to life from scratch. The scale of this procedural animation is wild once you zoom out: → Starts with 1 solitary spire in total darkness → Expands outward across 1,000+ generated skyscrapers → Illuminates 50,000+ warm grid windows in a radial wave → Renders a full cyberpunk skyline in 300 frames The best part? Every single building, street grid, and light pattern follows pure algorithmic rules—zero manual placement.
Your agent can spend all night repeating the same mistake Prompt, Context, Loop, and Harness Engineering Find the failing layer and learn exactly what to fix The four-layer playbook for agents that finish the job ↓
2
3
22
1,028
1 signal. 12 seconds. An entire city brought to life from scratch. The scale of this procedural animation is wild once you zoom out: → Starts with 1 solitary spire in total darkness → Expands outward across 1,000+ generated skyscrapers → Illuminates 50,000+ warm grid windows in a radial wave → Renders a full cyberpunk skyline in 300 frames The best part? Every single building, street grid, and light pattern follows pure algorithmic rules—zero manual placement.
Your agent can spend all night repeating the same mistake Prompt, Context, Loop, and Harness Engineering Find the failing layer and learn exactly what to fix The four-layer playbook for agents that finish the job ↓
2
3
22
1,028
Dain retweeted
12 STRANDS. 12 SECONDS. PICK ONE. TRY NOT TO LOSE IT. That’s the entire challenge. At first, it looks like a single chunk of chrome. Then the gaps open—and suddenly you’re not so sure where anything goes. 1 strand slips behind another. 2 reflections look like the same edge. The loop you were watching disappears into the knot. No cuts. No new pieces. All 12 strands are still there. And just when the shape starts making sense, it folds back into itself. Watch once for the reveal. Watch twice to see where you lost your strand. UNBOUND / DAIN
3
1
27
2,781
12 STRANDS. 12 SECONDS. PICK ONE. TRY NOT TO LOSE IT. That’s the entire challenge. At first, it looks like a single chunk of chrome. Then the gaps open—and suddenly you’re not so sure where anything goes. 1 strand slips behind another. 2 reflections look like the same edge. The loop you were watching disappears into the knot. No cuts. No new pieces. All 12 strands are still there. And just when the shape starts making sense, it folds back into itself. Watch once for the reveal. Watch twice to see where you lost your strand. UNBOUND / DAIN
3
1
27
2,781
Dain retweeted
18 LAYERS. ONE SHAPE YOU CAN’T SEE UNTIL THEY MOVE. At first, it looks like light trapped inside a solid block. Then the layers separate. Each one carries a different curve. Together, they form something none of them can show alone. I built this with code: → 18 transparent layers. → Amber light running through the structure. → 12 seconds from closed to open and back. Watch the middle as the glass pulls apart. That’s where the shape reveals itself.
6
2
37
969
18 LAYERS. ONE SHAPE YOU CAN’T SEE UNTIL THEY MOVE. At first, it looks like light trapped inside a solid block. Then the layers separate. Each one carries a different curve. Together, they form something none of them can show alone. I built this with code: → 18 transparent layers. → Amber light running through the structure. → 12 seconds from closed to open and back. Watch the middle as the glass pulls apart. That’s where the shape reveals itself.
6
2
37
969
Dain retweeted
1 WRONG PRICE. 4 PERFECT DOCUMENTS. 0 CHECKS AGAINST THE SOURCE. Imagine the price is $500. Your AI finds an old file that says $50. Then the workflow gets to work: → Email: $50. → Invoice: $50. → Spreadsheet: $50. → Report: $50. Everything matches. You’re still missing $450 per sale. Repeat that across 100 sales: $45,000 underbilled. No crash. No warning. Just a beautifully organized mistake. You ask another agent to check the invoice against the spreadsheet. It passes. Of course it passes. Both copied the same wrong number. I built this animation to show that chain reaction: 1 unchecked input spreading through 160 elements. Watch the red spread. Then ask the question that matters: Did your agents check the source… or just agree on the mistake?
5
3
41
9,874
1 WRONG PRICE. 4 PERFECT DOCUMENTS. 0 CHECKS AGAINST THE SOURCE. Imagine the price is $500. Your AI finds an old file that says $50. Then the workflow gets to work: → Email: $50. → Invoice: $50. → Spreadsheet: $50. → Report: $50. Everything matches. You’re still missing $450 per sale. Repeat that across 100 sales: $45,000 underbilled. No crash. No warning. Just a beautifully organized mistake. You ask another agent to check the invoice against the spreadsheet. It passes. Of course it passes. Both copied the same wrong number. I built this animation to show that chain reaction: 1 unchecked input spreading through 160 elements. Watch the red spread. Then ask the question that matters: Did your agents check the source… or just agree on the mistake?
5
3
41
9,874
Dain retweeted
Your AI remembered the plan. Unfortunately, it was yesterday’s plan. Monday: “We launch Friday.” Tuesday: “Move the launch to Monday.” Both messages are saved. Both can show up in a search. Only one should drive the next action. That’s the problem this animation explores: → Keep the original decision. → Mark it as superseded. → Link it to the update. → Use the current instruction. Imagine a teammate who remembers every meeting but misses every correction. Giving them a bigger notebook won’t fix that. AI memory needs a way to say: “I remember this. It no longer applies.”
Stop letting unfinished Opus 5.5 runs eat your budget Set up the harness before your next long task Save progress, verify results, reuse the workflow The full 7-layer setup below ↓
3
60
7,737
Your AI remembered the plan. Unfortunately, it was yesterday’s plan. Monday: “We launch Friday.” Tuesday: “Move the launch to Monday.” Both messages are saved. Both can show up in a search. Only one should drive the next action. That’s the problem this animation explores: → Keep the original decision. → Mark it as superseded. → Link it to the update. → Use the current instruction. Imagine a teammate who remembers every meeting but misses every correction. Giving them a bigger notebook won’t fix that. AI memory needs a way to say: “I remember this. It no longer applies.”
Stop letting unfinished Opus 5.5 runs eat your budget Set up the harness before your next long task Save progress, verify results, reuse the workflow The full 7-layer setup below ↓
3
60
7,737