Software Engineer and Tech Lead building full-stack products. Opinions are my own. https://xitter.gq/t.co/QLtcgRIYBb
Berlin, Germany
Joined October 2007
- Tweets1.3K
- Following258
- Followers417
- Likes688
Maximilian Fellner retweeted
Current LLMs aren't truly creative because they haven't implemented the "Formal Theory of Fun and Creativity" (2008) yet. See "Driven by Compression Progress: A Simple Principle Explains Essential Aspects of Subjective Beauty, Novelty, Surprise, Interestingness, Attention, Curiosity, Creativity, Art, Science, Music, Jokes" arxiv.org/abs/0812.4360 Tweet: xitter.gq/SchmidhuberAI/status/1…
Oxford researchers argue that LLMs can never invent anything.
It is mathematically impossible.
They published a paper called “Theory Is All You Need" and it argues against the claim that computational models can generate genuine novelty or new knowledge.
They analyzed the limits of generative ai, and the results are a brutal reality check for the idea that ai will replace human decision making under uncertainty.
Here is why AI is stuck and human cognition wins:
backward-looking vs forward-looking.. llms are probability machines that look backward at existing data. human cognition is forward-looking and capable of generating genuine novelty. human cognition operates theoretically "top-down" rather than "bottom-up" from data.
the "data-belief asymmetry".. the researchers use the invention of "heavier-than-air flight" to illustrate this concept. an ai relies on data-based prediction, which is largely imitative. humans, however, use theory-based causal logic that allows them to hold beliefs that go beyond existing data.
the intervention gap.. humans don't just process information; we use theory to practically "intervene" in the world. we engage in directed experimentation to generate entirely new data. ai-based models are theory-free and place primacy on existing data and prediction.
tldr?
AI uses a probability-based approach to knowledge and ia largely imitative. It can process data and make predictions, but human cognition relies on theory-based causal reasoning.
The decades-old analogy comparing human minds and computers to mere "input-output" devices is fundamentally flawed.
Maximilian Fellner retweeted
Introducing GLM-5.3-Flash
- Leading capabilities at a highly competitive price
- Natively multimodal with a 1M-token context window
- A 320B-A18B model released under the MIT License
- Previously previewed as Ox Alpha, running entirely on Chinese AI chips
Blog: z.ai/blog/glm-5.3-flash
Available now across all official platforms:
Weights: huggingface.co/zai-org/GLM-5…
API: docs.z.ai/guides/llm/glm-5.3…
Coding Plan: z.ai/subscribe
ZCode: zcode.z.ai/en
Chat: chat.z.ai
AutoClaw: autoclaw.z.ai
Maximilian Fellner retweeted
I support open-source models distilling what commercial companies distilled for free from the entire internet. I published distillation for free in 1991 in Europe - this was copied in the US and in China (xitter.gq/SchmidhuberAI/status/1…)
We have information that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model.
To do this they developed a sophisticated internal platform to conduct large scale distillation against U.S. models, allowing them to quickly switch between multiple methods of access to avoid detection. Moonshot AI has also acquired GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models.
The United States strongly supports the free and fair development of AI, including a thriving competitive ecosystem that spans frontier models, specialized systems, open-source frameworks, and open-weight models. Legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem. However, large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable.
Maximilian Fellner retweeted
State of Local AI #1
———
In lieu of Fable ban.
Here’s the best LLMs of the week to run on your hardware.
—— 4-8gb vram/ram 500$
- Gemma-4-qat huggingface.co/unsloth/gemma… I had someone mention it’s very good for subagent stuff
—— 8-16gb vram/ram < 1k usd
- Gemma-12B huggingface.co/google/gemma-… without a doubt the smartest model of its size
—— 16-32gb Apple/Strix halo 1-2k usd
- Diffusion Gemma26B huggingface.co/google/diffus…
- on 1x 6000 it’s eating up to 600 tok/s
- smallest smart MoE we have
- lots of world knowledge
- easy to run
—— 32-96gb ram/vram (2-10k usd)
- nex-n2-mini huggingface.co/nex-agi/Nex-N… builds on qwen3.6-35B and seems to do really well
- qwopus-27B huggingface.co/Jackrong/Qwop… this model topped a lot of our benchmarks at local.ai
—— 384gb vram (10-50K usd)
- huggingface.co/MiniMaxAI/Min… 23B means it’s close to qwen3.6-27B per token, while also have a lot of specialisation.
- fast inference
- top open weight model on AA
—— 768gb-1TB
- huggingface.co/moonshotai/Ki…
Kimi has always been a top player here and their last model cuts speed and cost down by 30%
- great vision support
- first coder model by moonshot
———
Top models:
1. Qwen3.6-35B
2. Qwen3.6-27B
3. Step-3.7-Flash
4. Minimax-M3
5. Deepseek-v4-flash
———
Budget sweet spots:
#1 - 1K usd
Single 3090 / Mac mini / Intel arc b70 / AMD
- Qwen / Gemma
#2 - 5k usd
DGX Spark / Mac m5 max / 4x 3090
- qwen / Gemma step and deepseek flash
#3 - 12k usd
RTX Pro 6000 / Mac Ultra / 2x Spark / 8x 3090
Ds4-flash / step-3.7-Flash and above
#4 - 24k usd
2x 6000 / 2x Mac Ultra / 4x Spark / Mix
Same as above
#5 - 50k usd
4x 6000 / 4x Max Ultra / 12x Spark / 2 H100
Minimax-m3 / nex-n2-pro / step-3.7-flash
#6 - 100k usd
GB300 station / 8x 6000 / 4x H200 / Mix
GLM-5.2 / Kimi-K2.7
———
Let’s keep the Internet free thanks for reading
🤖 Made with AI
Maximilian Fellner retweeted
Following the awesome @openclaw meetup in Vienna yesterday, who's up for an Open Claw meetup in Berlin in a few weeks (e.g. March 10 or 12)?
1) "like" if you like it
2) + retweet if you love it
3) + add a comment if you're game so we can keep you posted
Maximilian Fellner retweeted
Europe, it's time for action! 🇪🇺
Join our petition for an EU-Inc:
A pan-European startup entity
Goal:
- fix investments
- fix europe-wide employment
- fix stock options
- unite the European startup ecosystems
We got 6 weeks to make this real!
Sign here: eu-inc.org
Maximilian Fellner retweeted
Microsoft's new Kosmos-1 is incredible.
It's a new Multimodal Large Language Model (MLLM).
Their model can understand images, text, images with text, OCR, image captioning, visual QA.
It can even solve IQ tests.
Paper: arxiv.org/abs/2302.14045
Code: github.com/microsoft/unilm
Maximilian Fellner retweeted
So, what's next? github.com/zloirock/core-js/…
Maximilian Fellner retweeted
Maximilian Fellner retweeted
Do you like detective stories?
I just published the one about web performance! link.medium.com/ZUBkVT9XOcb
Slow website speed was already a fact when I joined my current company. I want to share my investigative story on finding performance criminals 🕵️♀️
#webperformance
Maximilian Fellner retweeted
Have you ever used Apollo CLient for local state management? With a new API, it is much easier now. Here is an article-tutorial about Storing local data with Apollo Client link.medium.com/w5xqpt0Ckab
Maximilian Fellner retweeted
What's next for hapi? A new community led effort is underway...
github.com/hapijs/hapi/issue…
Maximilian Fellner retweeted
We were asked to make a trailer for the #LVI @IEEESSP paper. Here it is ;)
"We turn #Meltdown around."
youtube.com/watch?v=baKHSXeI…
xitter.gq/jovanbulck/status/1237…
lviattack.eu/
cc @jovanbulck @danielmgmi @misc0110 @mlqxyz
We present Load Value Injection #LVI: a new transient-execution attack class; defeats defenses; turns around #Meltdown #Foreshadow #ZombieLoad #RIDL #Fallout to *inject* attacker data into victim loads. lviattack.eu cc @danielmgmi @mlqxyz @misc0110 @lavados @IEEESSP
Maximilian Fellner retweeted
Okay, my issue with `forEach`:
PROS
+ You can compose a reusable function for the handler. In the rare case you need that.
CONS
- No break
- No continue
- No yield
- No await
- Additional layer on the call stack
- Additional function allocation(s)
- More visual noise (IMO)
Maximilian Fellner retweeted
Is it a good idea to self-host Google Fonts? Probably yes. A very detailed piece by @tunetheweb.
tunetheweb.com/blog/should-y…
Also: a small shell script to download Google's web fonts
github.com/neverpanic/google…
Maximilian Fellner retweeted
1/ The Phoenix Project #ebook is *free* today (Dec 19) on @amazon & other bookseller sites.
If there’s someone you think should read it, now’s your chance. Retweet + tag your friends, colleagues, bosses.
Link: itrevolution.com/free-the-ph…
(A Thread) @RealGeneKim @kevinbehr @gspaff
