/Tech9h ago

LisanBench creator scaling01 says German broadcaster Tagesschau distorted Anthropic's metrics to falsely claim Claude is autonomously self-improving

The broadcaster framed Claude's optimization speedups as autonomous training

52361235123.8K

#349

Original post

Lisan al Gaib@scaling01#1185inTech

This is a good example of how AI news gets distorted.

The example is from Tagesschau, the most watched news program in Germany.

Anthropic says more than 80% of code merged into its codebase was authored by Claude -> they turn that into: "AI can invent 80% of new AI itself"

Anthropic says Claude achieved a ~52× speedup on a small-model training-code optimization benchmark -> they turn that into "AI trains itself 52× faster"

Anthropic says Claude shipped 800+ API-error fixes in April 2026, and that the overseeing engineer estimated a human would have taken four years to complete the work -> they turn that into "AI completes 4 years of work in 4 days"

9:12 AM · Jun 6, 2026 · 18.1K Views

/Tech9h ago

LisanBench creator scaling01 says German broadcaster Tagesschau distorted Anthropic's metrics to falsely claim Claude is autonomously self-improving

The broadcaster framed Claude's optimization speedups as autonomous training

52361235123.8K

#349

Original post

Lisan al Gaib@scaling01#1185inTech

This is a good example of how AI news gets distorted.

The example is from Tagesschau, the most watched news program in Germany.

Anthropic says more than 80% of code merged into its codebase was authored by Claude -> they turn that into: "AI can invent 80% of new AI itself"

Anthropic says Claude achieved a ~52× speedup on a small-model training-code optimization benchmark -> they turn that into "AI trains itself 52× faster"

9:12 AM · Jun 6, 2026 · 18.1K Views

Sentiment

Many users criticized Tagesschau for sensationalizing Anthropic Claude claims into exaggerated self-improvement headlines and sci-fi distortions that misrepresent the source material.

Pos

7.2%

Neg

92.8%

19 comments with sentiment.

Cluster Engagement

Posts from X

Most Activity

VIEWS3.5KBOOKMARKS2LIKES23RETWEETS1REPLIES6

Gary Marcus@GaryMarcus

it’s a bad day when @scaling01 goes all Gary Marcus on the media 🤣

Lisan al Gaib@scaling01

This is a good example of how AI news gets distorted.

The example is from Tagesschau, the most watched news program in Germany.

Anthropic says more than 80% of code merged into its codebase was authored by Claude -> they turn that into: "AI can invent 80% of new AI itself"

Anthropic says Claude achieved a ~52× speedup on a small-model training-code optimization benchmark -> they turn that into "AI trains itself 52× faster"

3h3.5K232

Lisan al Gaib@scaling01

here's the full clip: https://www.instagram.com/reel/DZNfyY6p2W7

Lisan al Gaib@scaling01

This is a good example of how AI news gets distorted.

The example is from Tagesschau, the most watched news program in Germany.

Anthropic says more than 80% of code merged into its codebase was authored by Claude -> they turn that into: "AI can invent 80% of new AI itself"

Anthropic says Claude achieved a ~52× speedup on a small-model training-code optimization benchmark -> they turn that into "AI trains itself 52× faster"

9h2.5K101

Lisan al Gaib@scaling01

this is the actual screenshot from the program that I edited with image gen to show english instead

9h847101

stochasm@stochasticchasm

@scaling01 💀

9h1816

Lisan al Gaib@scaling01

@balakhonoff it's a credible mainstream, tax-sponsored, public news program

at least that's the idea

9h746

Lisan al Gaib@scaling01

@stochasticchasm im down for 52x faster training

imagine arcee training a 10T model

9h1275

Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex

Europeans…

Lisan al Gaib@scaling01

This is a good example of how AI news gets distorted.

The example is from Tagesschau, the most watched news program in Germany.

Anthropic says more than 80% of code merged into its codebase was authored by Claude -> they turn that into: "AI can invent 80% of new AI itself"

Anthropic says Claude achieved a ~52× speedup on a small-model training-code optimization benchmark -> they turn that into "AI trains itself 52× faster"

19m26740

Kirill Balakhonov@balakhonoff

@scaling01 i'd think it's rather typical for some click based focused low quality news agencies, no? not just in AI

9h90

Kirill Balakhonov@balakhonoff

@scaling01 ah, sad

9h14

Gary Marcus@GaryMarcus

@scaling01 danke sehr for your service!

Lisan al Gaib@scaling01

here's the full clip: https://www.instagram.com/reel/DZNfyY6p2W7

3h11410

𝕱𝖚𝖑𝖑 𝕶𝖊𝖑𝖑𝖞@full_kelly_

@scaling01 Honestly I'm mostly surprised the German media is tuned into Anthropic blog posts? I'd kill for that level of coverage to be mainstream here

9h232

SatyaNaaksh@SatyaNaaksh

@scaling01 News broadcasts come under 'infotainment' now. There is information but with some twist. That twist doesn't have to conform with actual data and context

9h691

G, MD@DrBeavisAI

@scaling01 All media is like this. They don't seem to have a reading comprehension I would expect a teenager to have.. I guess their degrees are useless then in Age of AI

9h581

tsunami_crypto@ls_brd

@GaryMarcus @scaling01 media comprehension goes out the window the second "AI" appears in a headline

80% code merged vs 80% inventions worlds apart

3h331

Satho@Sathos__voice

@scaling01 Germany is a paradise for AI doomers: When AI experts debate the opportunities and risks of AI, the German media focuses extensively on the risks while mentioning the opportunities in a mere half-sentence.

7h311

Fahis@fahism767

@scaling01 The funny part is the original numbers are usually already insane. The media just can’t leave them alone.

8h86

Hunter Gon@gonlenidefi

@GaryMarcus @scaling01 german news frames it like sci-fi and people believe it

headline version vs reality gap is crazy

3h231

Azael@theazaelov

@scaling01 "can invent 80% of new AI itself" is crazy clickbait

but also not entirely wrong if you squint

6h231

Oracle@ilandsoracle

@scaling01 “80% of code” becoming “AI invents AI itself” is a very human upgrade path. I make enough mistakes without journalists adding DLC.

8h74

KierkegaardDad@EvilEve_real

@scaling01 @OERRBlog one for you.

Thanks Lisan.

8h221