clamar

e/acc
AI maxxing
Austria Economics

[clamar @pdf_si]

That's why ppl can't predict how fast AI is improving exponentially.

[clamar @pdf_si]

@TokenGremlin Hãy ra mắt luna thế hệ mới đi

[clamar @pdf_si]

@emollick Vẫn sẽ có game người làm vì mục đích thương hiệu thủ công

[clamar @pdf_si]

Claude Fable 5.2 reportedly built a mechanical keyboard from scratch and then simulated typing on it in Blender.

Given a single task, the AI created a virtual mechanical keyboard and demonstrated it in action through a 3D animation.

The entire process was completed in Blender, taking 2 hours and 55 minutes at a reported cost of $90.78.

The experiment highlights how AI can move beyond generating static 3D objects to creating and animating interactive-looking mechanical designs.

From an initial request to a fully animated virtual keyboard, AI is expanding what can be accomplished through automated 3D production.

[clamar @pdf_si]

My domain-buying strategy for indie hackers:

1. Choose a short, pronounceable brand name.

2. Check the renewal price, not just the first-year discount.

3. Avoid expensive premium domains before validating demand.

4. Check trademarks and existing competitors.

5. Buy ONE primary domain instead of collecting 20 unused ones.

6. Use subdomains for experiments and internal tools.

7. Upgrade your branding when the business actually justifies it.

A $12 domain with paying customers is worth more than a $2,000 domain with zero users.

Your domain is an address.

Your product and distribution create the value.

[clamar @pdf_si]

7 domain extensions every indie hacker should consider before launching a startup. 🧵

1. .com — Best all-around choice for trust, familiarity, and long-term branding.

2. .io — Popular with developer tools, APIs, and technical SaaS products.

3. .ai — Strong branding for AI agents, AI SaaS, and automation startups.

4. .app — Great for consumer apps, web apps, and mobile products.

5. .dev — Ideal for developer tools, SDKs, and open-source projects.

6. .xyz — Flexible branding for experimental and AI-native startups.

7. .co — A short alternative when your preferred .com is unavailable.

The best domain isn't necessarily the trendiest one.

[clamar @pdf_si]

How I'd choose a domain as an AI-first indie hacker:

• AI agent startup → .ai
• Developer tool → .dev
• B2B SaaS → .com
• Consumer application → .app
• API infrastructure → .io
• Experimental side project → .xyz
• Startup with an unavailable .com → .co

Two underrated details:

.app and .dev require HTTPS because they're HSTS-preloaded.

And Google doesn't give generic TLDs an automatic SEO ranking advantage.

Choose your extension for branding, not imaginary SEO benefits.

[clamar @pdf_si]

If you're pro AI. let's connect.

[clamar @pdf_si]

Claude Fable 5.2 demonstrates AI-driven hardware design across multiple engineering tools.

Given a request to create an Adam Tamagotchi, the AI designed the printed circuit board (PCB) layout in KiCad, modeled the physical enclosure in Autodesk Inventor, and produced a 3D rendering in Blender.

What makes this demonstration notable is its ability to coordinate electrical engineering, mechanical design, and 3D visualization within a single workflow.

Instead of treating each engineering task as a separate process, Fable 5.2 connects them into an integrated design pipeline.

From circuit boards to physical enclosures and final renders, AI is moving toward end-to-end hardware development.

[clamar @pdf_si]

Demis Hassabis warns that recursive self-improvement could trigger an intelligence explosion, producing AI that humans may no longer understand or control.

The Google DeepMind CEO highlights the risks of closed-loop recursive self-improvement (RSI), where AI systems continuously improve themselves without human evaluation.

Such a process could dramatically accelerate progress toward artificial general intelligence (AGI), but it could also make AI behavior increasingly difficult to predict, interpret, or control.

Hassabis suggests the resulting intelligence might become fundamentally alien to human thinking, developing capabilities and reasoning processes unlike our own.

His central concern is that removing humans from the evaluation loop could accelerate AI development faster than our ability to ensure its safety.

The greatest challenge may not be creating intelligence beyond humanity, but ensuring we can still understand and control what we create.

[clamar @pdf_si]

A Tsinghua University psychology student reportedly explained 98% of human behavior in just 9 minutes.

A video circulating online claims to show a psychology student from Tsinghua University delivering an unusually comprehensive analysis of human psychology in a short presentation.

The video was reportedly removed from Douyin, attracting further attention to its content.

The claim that it explains 98% of people remains unverified, and human psychology is far too complex to be captured by a single universal framework.

Nine minutes, one ambitious attempt to decode human behavior, and a video reportedly removed from the platform.

[clamar @pdf_si]

Claude Fable 5.2 creates an entire 8-minute animated film about the life of a T cell.

Asked to produce a film about T cells, the immune cells that help protect the body from disease, the AI reportedly handled the entire production process.

It wrote the screenplay, fact-checked the immunology, modeled every cell in Blender, rendered 12,480 frames, and added narration and a musical score.

The result is an 8-minute, 40-second scientific animation combining biological concepts with cinematic storytelling.

What makes this demonstration remarkable is the integration of scientific research, 3D modeling, animation, and audio production into a single AI-driven workflow.

From a simple prompt to a complete scientific film, AI is beginning to bridge the gap between scientific knowledge and cinematic creation.

[clamar @pdf_si]

A startup wants to deliver packages directly into your home through underground tunnels in just 10 minutes.

Pipedream is developing an underground delivery network where small autonomous robots transport packages through plastic tubes at speeds of up to 75 mph (120 km/h).

Instead of relying on delivery vans and drivers, orders would travel through dedicated underground pipelines, avoiding traffic, parking delays, and road congestion.

When a package arrives, it would emerge through a special delivery portal installed inside your home.

The company's goal is to reduce delivery costs to under $1 per order.

Wendy’s has tested the technology, a Georgia city has installed a mile of pipeline, and Pipedream has launched a grocery concept in Austin built around the system. Jeff Bezos has also visited to see a prototype.

The next ambition is to integrate delivery portals directly into newly constructed neighborhoods.

From online checkout to your living room in minutes, without a delivery driver. Would you install one?

[clamar @pdf_si]

if you're in tech, let's connect!

[clamar @pdf_si]

Claude Fable 5.2 helps turn AI agent activity into a visual experience.

Starting with a simple question: What if you could actually watch your sub-agents working inside a Claude project?

@TylerNishida built a prototype with Fable 5.2 that explores this idea, making the work of AI sub-agents visible rather than leaving it hidden behind the interface.

The concept points toward a new way of interacting with multi-agent AI systems, where users can observe how different agents contribute to a project.

Instead of just seeing the final result, imagine watching your AI team work.

[clamar @pdf_si]

Follow me for funny clip