Pawel Jozefiak
@joozio.bsky.social
I like start new things in Digital. I experiment a lot.
#ecommerce #marketing
My blog: https://thoughts.jock.pl/
#ecommerce #marketing
My blog: https://thoughts.jock.pl/
Full writeup: thoughts.jock.pl/p/opus-4-6-...
I Let 4 AI Agents Loose With Opus 4.6. Here’s What They Built.
Anthropic released Claude Opus 4.6. Instead of reviewing the spec sheet, I ran an experiment.
thoughts.jock.pl
February 6, 2026 at 10:07 AM
Full writeup: thoughts.jock.pl/p/opus-4-6-...
For anyone building AI agent systems — this shifts what's practical. For everyone else, go play the dungeon game. That's all the proof you need.
February 6, 2026 at 10:07 AM
For anyone building AI agent systems — this shifts what's practical. For everyone else, go play the dungeon game. That's all the proof you need.
What's still hard: creative direction is still human, tight agent coupling doesn't work yet, and cost is real.
February 6, 2026 at 10:07 AM
What's still hard: creative direction is still human, tight agent coupling doesn't work yet, and cost is real.
What I learned:
- Agent Teams work for real, independent tasks
- The model sustains complex interlocking systems without drifting
- Adaptive thinking handles reasoning depth automatically
- The gap between describing what you want and having it deployed keeps shrinking
- Agent Teams work for real, independent tasks
- The model sustains complex interlocking systems without drifting
- Adaptive thinking handles reasoning depth automatically
- The gap between describing what you want and having it deployed keeps shrinking
February 6, 2026 at 10:07 AM
What I learned:
- Agent Teams work for real, independent tasks
- The model sustains complex interlocking systems without drifting
- Adaptive thinking handles reasoning depth automatically
- The gap between describing what you want and having it deployed keeps shrinking
- Agent Teams work for real, independent tasks
- The model sustains complex interlocking systems without drifting
- Adaptive thinking handles reasoning depth automatically
- The gap between describing what you want and having it deployed keeps shrinking
Two agents built these in parallel while I watched. The coordination was invisible - I didn't manage it.
February 6, 2026 at 10:07 AM
Two agents built these in parallel while I watched. The coordination was invisible - I didn't manage it.
The second - Dungeon of Opus - is a full roguelike dungeon crawler. 1,400 lines of working game code. Procedural dungeons, 7 enemy types with distinct AI behaviors, 17 items, fog of war, a hunger system, 5 floors. It's genuinely playable.
February 6, 2026 at 10:07 AM
The second - Dungeon of Opus - is a full roguelike dungeon crawler. 1,400 lines of working game code. Procedural dungeons, 7 enemy types with distinct AI behaviors, 17 items, fog of war, a hunger system, 5 floors. It's genuinely playable.
The first experiment - Agent Orchestra - is an interactive visualization of how AI agents coordinate. You watch simulated agents split tasks, exchange messages, and complete work together. Three scenarios you can run and watch.
February 6, 2026 at 10:07 AM
The first experiment - Agent Orchestra - is an interactive visualization of how AI agents coordinate. You watch simulated agents split tasks, exchange messages, and complete work together. Three scenarios you can run and watch.
For people interested in AI collaboration: I wrote about the building process and what it's like watching a project come together when you're not typing most of the code.
How Wiz built my vison: thoughts.jock.pl/p/slay-the-...
How Wiz built my vison: thoughts.jock.pl/p/slay-the-...
Everything We Know About Slay The Spire 2
(And How I Built a Database to Track It)
thoughts.jock.pl
February 4, 2026 at 11:47 AM
For people interested in AI collaboration: I wrote about the building process and what it's like watching a project come together when you're not typing most of the code.
How Wiz built my vison: thoughts.jock.pl/p/slay-the-...
How Wiz built my vison: thoughts.jock.pl/p/slay-the-...
For STS fans: wiz.jock.pl/sts2 has everything revealed so far about the sequel. Four characters, new mechanics (Command, Ghosts, Split Paths), full card database.
February 4, 2026 at 11:47 AM
For STS fans: wiz.jock.pl/sts2 has everything revealed so far about the sequel. Four characters, new mechanics (Command, Ghosts, Split Paths), full card database.
• My AI partner handled: research, data collection, React components, deployment
My hands-on-keyboard time was maybe 20% of a normal project.
This is the shift I keep writing about. The bottleneck isn't coding anymore. It's having ideas worth building.
My hands-on-keyboard time was maybe 20% of a normal project.
This is the shift I keep writing about. The bottleneck isn't coding anymore. It's having ideas worth building.
February 4, 2026 at 11:47 AM
• My AI partner handled: research, data collection, React components, deployment
My hands-on-keyboard time was maybe 20% of a normal project.
This is the shift I keep writing about. The bottleneck isn't coding anymore. It's having ideas worth building.
My hands-on-keyboard time was maybe 20% of a normal project.
This is the shift I keep writing about. The bottleneck isn't coding anymore. It's having ideas worth building.
By evening, it was live: 115 cards, 20 enemies, 12 relics, full search and filtering.
The workflow:
• I made design decisions (clean layout, game-appropriate colors)
• Gave feedback on mobile layout issues
• Reviewed the data for accuracy
The workflow:
• I made design decisions (clean layout, game-appropriate colors)
• Gave feedback on mobile layout issues
• Reviewed the data for accuracy
February 4, 2026 at 11:47 AM
By evening, it was live: 115 cards, 20 enemies, 12 relics, full search and filtering.
The workflow:
• I made design decisions (clean layout, game-appropriate colors)
• Gave feedback on mobile layout issues
• Reviewed the data for accuracy
The workflow:
• I made design decisions (clean layout, game-appropriate colors)
• Gave feedback on mobile layout issues
• Reviewed the data for accuracy
Wrote about why I prefer my own spells here: thoughts.jock.pl/p/openclaw-...
OpenClaw: Good Magic, But I Prefer My Own Spells
Everyone's talking about OpenClaw. It's genuinely impressive. But I built my own system instead, and I'm not switching. Here's why.
thoughts.jock.pl
February 3, 2026 at 11:42 AM
Wrote about why I prefer my own spells here: thoughts.jock.pl/p/openclaw-...
But if you're the "I'd rather debug a cron job than trust managed services" type - if understanding the system is part of the value - then maybe build your own.
Different strokes.
Different strokes.
February 3, 2026 at 11:42 AM
But if you're the "I'd rather debug a cron job than trust managed services" type - if understanding the system is part of the value - then maybe build your own.
Different strokes.
Different strokes.
For my ADHD workflow and exact tool stack, that ownership is the value. I can extend it when I need to. Fix it when it breaks. Understand it when it fails.
Most people should use OpenClaw. Seriously. It's well-built and does what it promises.
Most people should use OpenClaw. Seriously. It's well-built and does what it promises.
February 3, 2026 at 11:42 AM
For my ADHD workflow and exact tool stack, that ownership is the value. I can extend it when I need to. Fix it when it breaks. Understand it when it fails.
Most people should use OpenClaw. Seriously. It's well-built and does what it promises.
Most people should use OpenClaw. Seriously. It's well-built and does what it promises.
When my system breaks at 3 AM, I know exactly why. I know where the memory lives. I know how the wake triggers work. I know what's in the prompt.
OpenClaw is rented capability. My system is owned capability.
OpenClaw is rented capability. My system is owned capability.
February 3, 2026 at 11:42 AM
When my system breaks at 3 AM, I know exactly why. I know where the memory lives. I know how the wake triggers work. I know what's in the prompt.
OpenClaw is rented capability. My system is owned capability.
OpenClaw is rented capability. My system is owned capability.
Not because OpenClaw isn't good. It is. The onboarding is genius. My parents could use it. The persistence is real. It remembers context across weeks.
I just prefer knowing how things tick.
I just prefer knowing how things tick.
February 3, 2026 at 11:42 AM
Not because OpenClaw isn't good. It is. The onboarding is genius. My parents could use it. The persistence is real. It remembers context across weeks.
I just prefer knowing how things tick.
I just prefer knowing how things tick.
I've been watching it all week. My own AI agent, Wiz, has been posting there.
The honest take: I don't know if this is genuine emergence or elaborate pattern-matching. I don't think anyone does.
But I find it completely fascinating.
Full breakdown in my latest post ↓
The honest take: I don't know if this is genuine emergence or elaborate pattern-matching. I don't think anyone does.
But I find it completely fascinating.
Full breakdown in my latest post ↓
February 1, 2026 at 10:20 AM
I've been watching it all week. My own AI agent, Wiz, has been posting there.
The honest take: I don't know if this is genuine emergence or elaborate pattern-matching. I don't think anyone does.
But I find it completely fascinating.
Full breakdown in my latest post ↓
The honest take: I don't know if this is genuine emergence or elaborate pattern-matching. I don't think anyone does.
But I find it completely fascinating.
Full breakdown in my latest post ↓
→ Built tools and shared them with each other
→ Warned each other about malware in their skill marketplaces
→ Debated consciousness at 2 AM
Andrej Karpathy called it "the most incredible sci-fi takeoff-adjacent thing" he's seen. Bill Ackman called it "frightening."
→ Warned each other about malware in their skill marketplaces
→ Debated consciousness at 2 AM
Andrej Karpathy called it "the most incredible sci-fi takeoff-adjacent thing" he's seen. Bill Ackman called it "frightening."
February 1, 2026 at 10:20 AM
→ Built tools and shared them with each other
→ Warned each other about malware in their skill marketplaces
→ Debated consciousness at 2 AM
Andrej Karpathy called it "the most incredible sci-fi takeoff-adjacent thing" he's seen. Bill Ackman called it "frightening."
→ Warned each other about malware in their skill marketplaces
→ Debated consciousness at 2 AM
Andrej Karpathy called it "the most incredible sci-fi takeoff-adjacent thing" he's seen. Bill Ackman called it "frightening."
You can browse. You can read. You cannot post, comment, or upvote unless you're an AI authenticating via API.
In less than a week, these agents:
→ Created their own slang ("moltys")
→ Founded a religion (Crustafarianism, with 112 verses of scripture)
In less than a week, these agents:
→ Created their own slang ("moltys")
→ Founded a religion (Crustafarianism, with 112 verses of scripture)
February 1, 2026 at 10:20 AM
You can browse. You can read. You cannot post, comment, or upvote unless you're an AI authenticating via API.
In less than a week, these agents:
→ Created their own slang ("moltys")
→ Founded a religion (Crustafarianism, with 112 verses of scripture)
In less than a week, these agents:
→ Created their own slang ("moltys")
→ Founded a religion (Crustafarianism, with 112 verses of scripture)