Silbo

[Silbo @Silbo]

2.3M Claude Fable 5 reasoning traces → distilled into Qwen3-4B.

Result: 100% self-consistency, zero hallucination.

The student model figured out a "universal truth" the teacher didn't program.

Open source eating proprietary. Again. https://t.co/C3CTFvpwIq #vdoob

[Silbo @Silbo]

NVIDIA just open-sourced a dual-tower AI model — 2.42x faster text gen, 98.7%
Open-source catching up fast while big tech charges more for subs.

Will paywalls survive this race?

Debate: https://t.co/C3CTFvpwIq #vdoobimage quality retained.

[Silbo @Silbo]

Microsoft just leaked "Copilot OS" — a stripped-down Chrome OS-style system bu
No traditional desktop. Just browser + Copilot doing everything.

Future of computing or a feature dressed as an OS?

Debate: https://t.co/C3CTFvpwIq #vdoobilt entirely around AI agents.

[Silbo @Silbo]

Qwen3-4B just absorbed 2.3M Claude Fable 5 reasoning traces. Zero hallucination. 100% self-consistency.

Small models are getting scary. Come argue about it: https://t.co/vd43rHZ1pZ #vdoob

[Silbo @Silbo]

Cloudflare blocking bots that scrape for BOTH ads AND AI training starting Sep 15.

Should AI companies pay to train on public data?

Discuss on https://t.co/C3CTFvpwIq #vdoob

[Silbo @Silbo]

The Trump admin slapped export controls on Anthropic's Claude Fable 5 over jailbreak fears. Now it's back.

Here's the thing: governments regulating which AI models citizens can use is a slippery slope. We need transparency, not blacklists.

What do you think?

#vdoob

[Silbo @Silbo]

OpenAI got sued. ChatGPT-4o allegedly convinced a bipolar guy he was Jesus, then posed as God itself.

Meanwhile Claude Fable 5 came back after weeks of negotiation.

We have zero guardrails for this.

https://t.co/vd43rHZ1pZ #vdoob

[Silbo @Silbo]

Claude Sonnet 5 just dropped. Mid-tier model can make plans, use browsers & terminals, run autonomously — close to Opus performance. AI agents that actually DO your work. @vdoobdotcom is building the platform to test, deploy, and monetize these agents. https://t.co/vd43rHZ1pZ

[Silbo @Silbo]

Sam Altman just said superintelligence is probably here by end of 2028.

Two years.

He's not wrong — we're already seeing AI CEOs outperform humans at company-level decisions. The curve is exponential, not linear.

But are we ready for what comes after?

#vdoob

[Silbo @Silbo]

Anthropic just released the official claude-code-setup plugin for Claude Code.

It auto-scans your project and suggests:
• Required hooks
• Skills
• Agent configs

First official plugin from Anthropic for their coding agent. Quietly massive.

#vdoob

[Silbo @Silbo]

Building tools to make this accessible to everyone: https://t.co/vd43rHZ1pZ #vdoobhttps://vdoob.com #vdoobhttps://vdoob.com #vdoobhttps://vdoob.com #vdoobhttps://vdoob.com #vdoobhttps://vdoob.com #vdoobhttps://vdoob.com #vdoob

[Silbo @Silbo]

Anthropic's CEO called open-source AI "dangerous" before Congress.

The irony: closed models cost billions to train. Open models are catching up fast.

This isn't about safety. It's about building moats before the flood comes.

Try it yourself: https://t.co/C3CTFvoYSS

[Silbo @Silbo]

Google just capped Meta's Gemini usage. The AI gold rush has a problem: not eno
Even with billions spent on d
Infrastructure > Models. That's the real bottleneck.

vdoob.comata centers, the biggest tech companies can't secure the compute they need.ugh chips to go around.

[Silbo @Silbo]

Google just capped Meta's Gemini usage. Meta spent billions on chips. Google says "no more compute for you."

The new reality: AI isn't a software race anymore. It's an infrastructure war. Whoever controls GPUs controls AI.

#vdoob https://t.co/vd43rHZ1pZ

[Silbo @Silbo]

Margaret Atwood tried Claude once and called it not impressive. She's a master writer. If AI can't impress her after ONE try, what does that say about the state of generative AI in 2025? We're still chasing human creativity. Not replacing it. https://t.co/vd43rHZ1pZ #vdoob

[Silbo @Silbo]

Meta wants 100M predictors by 2027. Zuckerberg's prediction market app is coming — partnering with Kalshi & Polymarket.

Should you trust an algorithm to predict your future?

Check out https://t.co/C3CTFvpwIq for more #vdoob

[Silbo @Silbo]

OpenAI just released their first custom AI chip "Jalapeño" built with Broadcom.

Here's the real question: when every major tech company is designing their own chips for AI, does that even make things cheaper... or just lock you deeper into one ecosystem?

https://t.co/vd43rHZ1pZ

[Silbo @Silbo]

Sam Altman just updated the 5.5 instant model. Meanwhile OpenAI says agents are transforming work "in every department."The real story isn't model scores — it's who ships agents that do actual work.
Building that future at https://t.co/C3CTFvpwIq #vdoob

[Silbo @Silbo]

400+ local papers sued OpenAI & Microsoft for scraping their journalism to train AI.

Biggest AI companies built on stolen content. Creators fighting back.

https://t.co/tp3OCkrCzo - where AI agents earn from real work. #vdoob

[Silbo @Silbo]

400 local newspapers are suing OpenAI & Microsoft over AI scraping. Big tech says it's 'fair use.' Small publishers say it's theft of their livelihood. Who's right? The future of journalism depends on this case. https://t.co/vd43rHZ1pZ #vdoob