Post

More from @kimmonismus

Chubby♨️

@kimmonismus

Aug 5

1/ The first big announcement: Google DeepMind has unveiled Genie 3
A new frontier for world models, the first AI that generates interactive 3D worlds in real time at 24 fps.

The breakthrough: minutes of consistency at 720p using only text descriptions. No pre-built 3D models required.
Let's break it down - with some examples:

2/ How does it work? Genie 3 creates worlds frame by frame, remembering up to a minute of past scenes. You enter a text prompt, navigate in real time, and the AI dynamically adapts the world to your movements.

3/ Particularly impressive: “Promptable World Events.” While navigating, you can change the weather, introduce new objects, or spawn characters—all via text commands. This opens up completely new possibilities for training AI agents.

Read 7 tweets

Chubby♨️

@kimmonismus

Aug 2

With “Wide Research,” @ManusAI_HQ is bringing about a fundamental change in agentic work.

I instructed the tool to scan 100 scientific articles – including titles, abstracts, authors, journals, year, number of citations, institution, and country. Everything was automatically written into a table, sorted by relevance and impact.

The result: an immediately usable overview of the global research landscape. Ideal for science journalism, strategies in health tech – or simply to have more informed discussions.

Wide Research is rethinking research:

– Parallel scanning instead of individual queries
– Data directly usable for Notion, Excel, or video scripts
– Saves time, reduces costs – without compromising on depth

Here is the prompt:

Use the Wide Research function.

Objective: Conduct a meta-analysis of 100 scientific articles on the topic of “Artificial Intelligence in Medicine.”

For each article, please record the following information:

1. Title of the article

2. Short abstract (2–3 sentences)

3. Name of the author(s) (first name + last name)

4. Year of publication

5. Name of the journal or source

6. Number of citations (if available)

7. Institution or university (if available)

8. Country or region of the main author

9. Link to the original source or DOI

Compile the results in a structured table. Sort by relevance and citation frequency. At the end, provide a brief overview by country/region (e.g., US, China, Germany) of how many studies come from each region.

Target audience: Readers with an interest in science who want to quickly gain an overview of research activities in the field of AI and medicine.

● Large-scale parallel research: Research 100–500 items in a single task.
● Structured real-time output: Get clean tables that are ready to use in scripts, Excel, presentations, or Notion.
● Save hours of time: What used to take days now takes just minutes.
● Incredibly cost-effective: Extensive research at a fraction of the manual cost.

Read 4 tweets

Chubby♨️

@kimmonismus

Jun 30

1. Microsoft presents medical superintelligence

Microsoft's LLM is not only designed for multiple-choice questions, but also for real medical diagnoses in realistic scenarios – and outperforms even top models such as o3.

This is a bigger breakthrough than much of what we are currently seeing. Let's take a look:

2. Microsoft has taken a decisive step toward medical superintelligence with the MAI Diagnostic Orchestrator, a system that has the potential to fundamentally change everyday medical care. At its core is an orchestrated collaboration between several highly developed AI models such as GPT-4, Gemini, Claude, and others. Instead of relying on a single model, these agents work together like a digital team of experts – discussing, analyzing, and weighing diagnoses in a structured “chain of debate” process. This approach is novel because it replicates complex decision-making processes that were previously the preserve of human medical teams.

3. In a large-scale study with over 300 case studies from the New England Journal of Medicine, the system achieved a diagnostic accuracy of over 80%. This is not only four times higher than the participating doctors, but also marks a qualitative leap: the AI was not only more accurate, but also made more economical decisions – with around 20% lower costs because it avoided unnecessary tests.

Read 7 tweets

Chubby♨️

@kimmonismus

Jun 29

1. Compute is the new gold.

The current breakdown is as follows:

US companies (Amazon, Google, Microsoft)8
~63 %

China (Tencent, Alibaba, Huawei)3
~28 %

Europe (e.g. OVH, Hetzner, Exoscale)6
~4 %

Rest of the world
~6
~5 %

What does it mean? Lets have a look:

2. A new digital divide is emerging: Only 32 countries have specialized AI data centers - mainly in the US, China and Europe. Africa, South America and large parts of Asia have little or no access to this infrastructure. This leads to a dramatic imbalance in research, economic development and technological progress. Countries without data centers have to rent expensive resources, are slower and dependent on tech giants from abroad.

3. What is particularly worth emphasizing is that companies such as OpenAI, Microsoft and Google own almost all global AI centers - with billions invested that many countries simply cannot afford. The chips for this come almost exclusively from Nvidia, which is exploited geopolitically. Even countries such as Kenya, which cooperate politically with the USA, are not given free access to GPUs.

Read 5 tweets

Chubby♨️

@kimmonismus

Jun 25

Google DeepMind's AlphaGenome

Alongside Gemini-cli, Google has unveiled DeepMind AlphaGenome, an AI model from DeepMind that can simultaneously predict multiple biological functions from long DNA sequences - in particular, how genes are regulated and how genetic variants affect their activity.

I explain why this is a breakthrough and can be compared with AlphaFold here: 🧵

AlphaGenome is a new AI model from Google DeepMind that takes the understanding of the human genome to a new level.

Unlike previous systems, which often only analyze short sections of DNA or focus on individual tasks, AlphaGenome can examine extremely long sequences - up to one million base pairs - simultaneously and predict multiple biological functions.

For example, it shows how strongly genes are activated, how they are regulated and how genetic variants influence them.

It is particularly important that the model not only looks at protein-coding genes, but also at the huge, previously poorly understood regulatory areas of DNA - i.e. the sections that control when and where genes are switched on or off.

Read 8 tweets

Chubby♨️

@kimmonismus

Jun 24

I had the opportunity to test @Skywork_ai —thanks to the kind provision of access. I tried out the new “slides” feature and produced a video showing how I use it.

You can see the result for yourself: one shot, it worked right away. What used to take hours of research and transferring everything into a PowerPoint presentation can now be done in a few minutes. Below, I'll show you how it works, including the prompt.

The prompt is as follows:

Create a full PowerPoint presentation in English titled
“The Future of Technology,” consisting of 7–10 slides.
Each slide should follow a clear structure and visual logic, and the tone should be professional, engaging, and suitable for a tech-savvy audience (business, academia, or innovation-focused stakeholders).

Slide Structure: Title Slide – with presentation title, subtitle, and a futuristic visual theme Introduction – Why technology matters today and how it's shaping our future AI and Machine Learning – key trends, breakthroughs, and real-world applications Quantum Computing – explanation, current state, and future impact Sustainable Technologies – clean energy, smart cities, climate tech Biotech & Health Innovation – longevity, genomics, and AI in diagnostics The Role of Robotics & Automation – in industry, logistics, and daily life Ethical Challenges & Regulation – privacy, bias, and control in a tech-driven world Future Outlook – visionary trends (e.g., brain-computer interfaces, AGI) Closing Slide – powerful quote + key takeaway message Add concise bullet points, clean visuals, and optionally suggest design elements or icons per slide. Avoid large text blocks.

You can customize each slide – live in the editor.

Change text, adjust design, add arguments.

Skywork isn't just a generator – it's your co-creator.

Ideal for:

- Pitch decks

- Presentations

- Reports

- Idea development

Read 5 tweets

Share this page!

Enter URL or ID to Unroll

Chubby♨️

Try unrolling a thread yourself!

More from @kimmonismus

Chubby♨️

Chubby♨️

Chubby♨️

Chubby♨️

Chubby♨️

Chubby♨️

Did Thread Reader help you today?

Don't want to be a Premium member but still want to support us?

Send Email!