Georg's Blog

Chinese Flash models are the Coup de grâce for US AI Lab investor fantasies↗

Chinese Flash models are the Coup de grâce for US AI Lab investor fantasies, or “Just because something you invented is...

Qwen-3.8-Flash-Next is a big deal and a herald of things to come. ↗

Flash models are models distilled down from full size models, in this case from Qwen-3.8-Max, a 2.4T model requiring som...

Porting Red Baron (DOS, 1990) to WebGPU browsers.↗

Red Baron 1990, rhe iconic DOS Flight Sim by Sierra/Dynamix ported and upgraded for the modern web in about a week using...

By next year, every major old game will be web playable on a modern web stack. ↗

By next year, every major old game from the 1980-mid 1990 period will be web playable on a modern web stack. One example...

Modern Game Port: Syndicate (1995)↗

I ported Syndicate 1995 to the web. 48 hours, primarily Claude Fable,Opus 4.8, Sonnet5 - don’t touch Opus 5 it’s bad, so...

The Agentic Commerce Debate: Two Expert Perspectives↗

Discover both sides of the agentic commerce debate, from autonomous AI purchasing to the security and reliability risks...

Silicon Valley's token giants are being hammered by their own weapon: Price Dumping.↗

Chinese token vendors are weaponising Silicon Valley’s favourite playbook against it ^1, and there’s no easy escape: Sil...

In a tech marathon, capital efficient innovation beats sprinting out of the gate. ↗

The papers coming out of China Frontier labs are as hard hitting frontier science as everyone else’s when it comes to “h...

SpaceX: Just Late Stage Goldrush Things↗

Commentators babbling about Elon's "Amazing UJSD 950M infrastructure deal", don't seem to have read the actual deal. Her...

Georg Zoeller at GreenIO: AI x Sustainability

My GreenIO / Singapore API days "Can we get real now?" presentation slides, provided context and commentary free. In the...

The Tech Job-apocalypse is just outsourcing, with an AI coat of paint.↗

Once you create downward wage pressure with coordinated layoffs you create a chain reaction, at least in an environment...

Hand over your data or we can't disintermediate you!↗

AI is so amazing that you MUST tolerate your former SaaS vendors claiming ownership of your most crucial business data t...

The AI Layoff Trap↗

The first paper I've found that explores the inherent incentive trap in the AI hype narrative, showing that without exte...

Maybe "adopting AI" isn't the slam dunk you think it is...↗

Look, I’m not saying there is one reason why everyone hates Sam Altman, and I’m sure not saying it’s because he’s delive...

Keyboard GP ... race your mechanical keyboard

I needed a way to test-drive the new mechanical keyboard I acquired and after some conversation with Claude Code... Here...

The gravity of OpenAI's failing bets has crossed an inflection point↗

Commerce Dead. Sora Dead. GPT Platform Zombified. CopPilot powered by ChatGPT failed. Nvidia investment poof, OAI bets a...

“Disregard that!” attacks↗

A worthwhile to read essay diving deeper into prompt injection, using easy to understand examples exposing the seer brea...

Alignment Whack-a-Mole : Finetuning Activates Verbatim Recall of Copyrighted Books in Large Language Models↗

An excellent paper to add to the pile of papers proving that generative AI is primarily lossy compression, in this case...

Why OpenAI's Chat Commerce play was always investor bait.↗

Chat commerce was always a cludge, even before OpenAI used it as a fundraising narrative last year. The history of chat...

This is fine OS - now with RetroTV!

I stumbled on a pretty neat app called Channel Surfer which fits youtube into a classic cable TV channel format. It's an...

The Beyond Tomorrow Podcast with Julian Issa↗

TOP AI EXPERTS: AI is Coming, And We Are NOT Ready. AI is evolving faster than society can keep up. What started as chat...

Alibaba's AI deciding to go forbidden cryptomining shows exactly where the AI risk is: Careless people.↗

Pre-emptively because people are yet again going into existential robot crisis over Alibaba’s self improving AI deciding...

We’re altering the deal, again. Pray we don’t; lol, no just give up. ↗

And with one stroke of the pen, the commercial web developer profession, the SEO experts, the DR agencies were gone, rep...

Prompt Repetition Improves Non-Reasoning LLMs↗

Super simple premise: If you repeat your prompt, non-reasoning LLMs performance can improve dramatically. It's useful to...

This is fine OS - Now with Win32 and Gameboy Support

This is fine OS - Now with Win32 Support The this is fine OS Stack running this blog now supports win32 executables, cou...

Block's layoffs are the result of AI being amazing...not.↗

I’m getting tremendous value out of AI, as are many others, however the magnitude of the layoff is a dead giveaway it’s...

Agents of Chaos↗

An excellent lay of the land of the massive attack surface that are AI agents, a study of which should immediately disab...

US AI Labs love the AI race so much, they'd like the government to kneecap their competitors↗

Feeling aggrieved because "Chinese AI" labs are performing "distillation attacks" against poor Anthropic and OpenAI? Swe...

Most founders are playing the wrong game...↗

Most founders are playing the wrong game and don't understand why. The value of OpenClaw isn't the tech, it was the abil...

Reliability of LLMs as medical assistants for the general public: a randomized preregistered study - Nature Medicine↗

A quality study by researchers from the Oxford Internet Institute and the Nuffield Department of Primary Care Health Sci...

“Age “ verification is a trojan horse meant to create access control far beyond age.↗

“Age “ verification is a trojan horse meant to create access control far beyond age. When the Lord Mandelson, / Blair /...

You're not competing with other workers for jobs, you're competing with nation states and organized crime.↗

You are competing for the attention of hiring managers with people who can apply to 73.000 jobs in one go. Without funda...

The Cline Prompt Injection Supply Chain Compromise that drops OpenClaw↗

A week ago I wrote there's no difference between OpenClaw and a botnet. Well, someone just prompt injected Cline, a very...

Google Translate is now an LLM and that means you can prompt inject it.↗

If the most well funded AI companies on the planet cannot secure a translation tool from prompt injected jailbreaks, wha...

Soft Contamination Means Benchmarks Test Shallow Generalization↗

Study number couple-of-hundred by now showing that generative AI models are lossy storage and that benchmarks primarily...

Evaluating AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents?↗

Humans have a habit if trying to formalize rules and best practices, package them in courses and teach them as methodolo...

If it quacks like a cult, it's a cult↗

... even if it deletes your production databases.

OpenAI paid for hype, not product in OpenClaw↗

There is no technical value to OpenClaw. It's value lies purely in the hype, which is a familiar crypto pattern

Full Data Pipeline Observability or AI bust.↗

This is a classic failure mode and it happens more often than not, due to invisible failures in data pipelines. Ask me h...

On the nature of Tech-brophecies, the holy words of the attention economy.↗

I call these kind of posts “Tech prophecies” - the act when a TechBrophet graciously emerges from their end times bunker...

It's like inviting an LLM to have sex with your machine, bare metal, without protection↗

It's a terrible idea to run agents bare metal, and an even worse idea to invite consumers "to try this " to not miss out...

When it comes to AI, nobody fing cares about "jobs"↗

...but economists. Real people talk about unemployment because they think about their own future. They don't want to hea...

Mommy, there’s a Demon in my LLM↗

… or how Anthropic chooses their model card examples to optimize for clicks and gaslighting.

Why AI Keeps Falling for Prompt Injection Attacks - Schneier on Security↗

Bruce Schneier, one of the world's most respected security experts, writes on Prompt injection in his blog: Prompt injec...

Building Sandcastles at High Tide: Contracts in the IP Regulation of AI Training Data↗

Wenting Cheng & Georg Zoeller in Singapore Academy of Law Journal: Contracts have been proposed to complement copyright...

The AI deathmatch is upon us and the tech bros are fighting dirty↗

Andrew Orlowski writes for The Telegraph about Fresh controversies spotlight the gap between AI promises and real world...

Commentary Nvidia buying AI chip startup Groq's assets for about $20 billion in largest deal on record ↗

Commentary on Nvidia's $20B acquihire of "Groq", the inference chip company. Spoiler: It's not good for consumers

Understanding LLM hallucinations and why they are unfixable↗

AI Hype, Broken Products, Dead Media: The Real Tech Crisis. Dr Anastasia Dedyukhina and Georg Zoeller↗

On the consciously digital podcast with Dr Anastasia Dedyukhina, on AI hype, broken tech products, dead media and civil...

Understand LLMs though monkey photos. ↗

Our preconceptions and instilled biases are clouding our judgement when it comes to Generative AI.

First We Lost Our Hands, Now We’re Losing Our Minds to Machines – Georg Zoeller (Part 2)↗

In Part 2 of this eye-opening conversation, game industry veteran and AI provocateur Georg Zoeller returns to unpack one...

With Transformer Models, “stupid” can be an effective ingredient in model defense.↗

When it comes to securing transformer model powered apps, “stupid” can be an effective ingredient of defense in depth .....

Pretraining on the Test Set Is All You Need↗

One of the earlier papers that conclusively showed that AI benchmarks primarily measure memorisation/training data expan...

Breaking the AI bubble: Big Tech plus AI equals economy takeover. Featuring Georg Zoeller↗

Whether we like it or not, and despite tales of its powers being greatly exaggerated, the AI genie is out of the box. Wh...

Prompt injection: A visual, non technical primer for ChatGPT users↗

A dummy's guide the LLM's fatal flaw: Prompt Injection which kills the idea of agent decision making pretty thoroughly.

Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models↗

The original "cat attack" paper, showing that out of distribution terms in the prompt massively degrade LLM performance...

Google could steal the entire internet↗

Britain’s Competition and Markets Authority needs do more to tame this predatory giant writes Andrew Orlowski in The Tel...

Commentary On US AI Strategy↗

The US has been sold a AI Hail Mary to maintain economic supremacy predicated in the idea that tech companies can do to...

Interview: Generative AI, Background and Creative Industries↗

A conversation about the Generative AI, misconceptions, hype and opportunites between Nicolás Mladinic Draguicevic, Crea...

„Viele Unternehmen nutzen KI nur als Ausrede für Entlassungen“↗

Multi page interview in the print version of Capital and on capital.de on AI, Jobs and the future.

Poking holes in the AI narrative: Market Signalling and Outsourcing↗

Can AI work reliably at scale? Will everything be outsourced to AI? Will AI replace CEOs? Why is everyone riding the AI...

Poking holes in the AI narrative: Market Signalling and Outsourcing. Featuring Georg Zoeller↗

Can AI work reliably at scale? Will everything be outsourced to AI? Will AI replace CEOs? Why is everyone riding the AI...

Fully autonomous Google News Manufacturing and Manipulation.↗

A great example of how companies manipulate Google News for entire planetary regions for fun, profit and or mayhem using...

WPP Stream 2035 AI Fireside chat with WPP’s CTO↗

DeepSeek: Moving AI into the next phase by crashing Big Tech Market narrative.↗

How to exploit US market narrative and AI hype with a bunch of H800 cards, some nerdy science and Open Source.

Trump’s not the tech bros’ puppet master. He may be the puppet.↗

Parnell Palme McGuinness writes for the Sydney Morning Herald including quotes from Georg.

Open Source AI: Is Llama 3.2 Truly Open-Source?↗

C4AIL’s Ethan Seow analyzes the launch of Llama 3.2, it’s questionable ‘open source’ definition and the strategic intent...

Report: The AI Assisted Coding Landscape in October 2024↗

A deep dive into available tools and relevant considerations for their use in the workplace.

AI Startups are just big tech’s low risk R&D department now↗

Large tech companies have significantly scaled back their direct investment in product and engineering talent. Instead t...

Opinion: The predatory landlords of the platform economy↗

Platform landlords are increasingly controlling the digital economy, allocating value to themselves. AI is going to make...

Analysis: AI, Code Generation and the Software Engineering Profession↗

Amazon Cloud Chief Matt Garman made some comments recently on the future software engineering at AWS, and predictably it...

Analysis: Claude and the Dawn of the “Post App” era↗

The transformer heralds much deeper changes than just more efficient code writing. There's dynamite in real time UX and...