Georg Heiler

GitHub - tobilg/duckdb_featureserv: A lightweight RESTful geospatial feature server for DuckDB with duckdb-spatial support A lightweight RESTful geospatial feature server for DuckDB with duckdb-spatial support - tobilg/duckdb_featureserv

cool - feature server goes #duckdb #gis github.com/tobilg/duckd...

GitHub - complexity-science-hub/llm-in-a-box-template: Template to use genai for chatting and via api to accelerate research Template to use genai for chatting and via api to accelerate research - complexity-science-hub/llm-in-a-box-template

Use the API? github.com/complexity-s... this makes it easy

Kart Distributed version-control for geospatial and tabular data.

kartproject.org #vcs #geospatial

Why we still need small language models – even in the age of frontier AI Lean, locally run models can unlock huge benefits for public sector and compute-constrained environments

#ai #small #models www.turing.ac.uk/blog/why-we-...

great to see package managers mature (not hat maven/gradle) java dependency overrides are nice - but sometimes just required

feat: pypi-options.dependency-overrides by HernandoR · Pull Request #3948 · prefix-dev/pixi add dependency-overrides for overriding. only updated manifest for now related #3917 #3890

@prefix.dev awesome to see dependency overrides merged github.com/prefix-dev/p...

We are currently preparing a submission for JOSE and are looking for feedback but also more people who are interested in joining in and refining the template

GitHub - complexity-science-hub/llm-in-a-box-template: Template to use genai for chatting and via api to accelerate research Template to use genai for chatting and via api to accelerate research - complexity-science-hub/llm-in-a-box-template

Together with www.linkedin.com/in/aaron-cul... I have created a template for making #LLMs (from different vendors and even self hosted ones) easily accessible to researchers - including advanced document RAG with #docling. github.com/complexity-s...

YouTube video by fern The Hunt for the World's Most Dangerous Hackers

#fern #hacker #bears #apt www.youtube.com/watch?v=ZhfI...

People say “your imagination is the limit” to mean the possibilities are limitless, but I believe that for many people the phrase is more literally true: they really are limited by a lack of imagination more than anything else

Sogar Notstrom fehlt: Schlechte Sicherheitstandards in Rechenzentren des Bundes Ein Bericht des Bundesrechnungshofs wirft kein gutes Licht auf die Sicherheit der IT des Bundes. Nur ein Bruchteil der Rechenzentren erreiche Mindeststandards.

#state of #cyber for #germany www.heise.de/news/Bundesr... quite sad to read that emergency power supplies are not there for a large quantity of the data centers

nice to see DE tooling adapted in other verticals - github.com/michimussato... here for visualization rendering

Principle 3 is that Agent actions and planning must be observable. I love this principle—emphasis mine: We cannot ensure an agent is acting faithfully or diagnose problems if its operations are entirely opaque. Therefore, agent actions, and where feasible, their planning processes, must be observable and auditable. [...] Effective observability also means that the properties of the actions an agent can take—such as whether an action is read-only versus state-changing, or if it handles sensitive data—must be clearly characterized. This metadata is crucial for automated security mechanisms and human reviewers. Finally, user interfaces should be designed to promote transparency, providing users with insights into the agent’s “thought process,” the data sources it consulted, or the actions it intends to take, especially for complex or high-risk operations. Yes. Yes. Yes. LLM systems that hide what they are doing from me are inherently frustrating—they make it much harder for me to evaluate if they are doing a good job and spot when they make mistakes. This paper has convinced me that there’s a very strong security argument to be made too: the more opaque the system, the less chance I have to identify when it’s going rogue and being subverted by prompt injection attacks.

I particularly appreciated principle 3: Agent actions and planning must be observable

The Guardian launches Secure Messaging, a world-first from a media organisation, in collaboration with the University of Cambridge Secure Messaging is a new innovation for confidential story-sharing and source protection, underpinning the Guardian’s commitment to investigative journalism. The Guardian has published the open sourc...

Congratulations @lambda.bsky.social! Today @theguardian.com is launching a new way for whistleblowers to anonymously contact journalists, based on years-long research by Daniel and other colleagues. www.theguardian.com/gnm-press-of...

Build optimized websites quickly, focus on your content | Docusaurus An optimized site generator in React. Docusaurus helps you to move fast and write content. Build documentation websites, blogs, marketing pages, and more.

docusaurus.io

YouTube video by Data Council Write Less More: How Dagster Rebuilt Our Docs From the Ground Up

#dagster journey for refining the documentation explained by @pedramnavid.com www.youtube.com/watch?v=dVB5... #datacouncil

YouTube video by Data Council Liberate Analytical Data Management with DuckDB

story of #duckdb www.youtube.com/watch?v=o53o... @hannes.muehleisen.org awesome talk!

YouTube video by Data Council Building Blocks: Advanced Semantic Data Model Layering

#malloydata #datacouncil #sql #semanticlayer www.youtube.com/watch?v=DZkX...

YouTube video by Data Council Building InfluxDB 3 Core: Real-Time Columnar DB & Data Processor On Object Storage

#influxdb #datafusion #datacouncil www.youtube.com/watch?v=8A4R...

Brett Adcock on X: "Japan's Sakana AI dropped Darwin Gödel Machine, a self-improving agent that can modify its code to boost performance On SWE-bench, DGM improved its performance from 20.0% to 50.0%, while on Polyglot, it increased its success rate from 14.2% to 30.7% https://t.co/Y7IZVjTQZa https://t.co/axNrZZvt7M" / X Japan's Sakana AI dropped Darwin Gödel Machine, a self-improving agent that can modify its code to boost performance On SWE-bench, DGM improved its performance from 20.0% to 50.0%, while on Polyglot, it increased its success rate from 14.2% to 30.7% https://t.co/Y7IZVjTQZa https://t.co/axNrZZvt7M

#darwin #gödel #ai x.com/adcock_brett...

YouTube video by DuckDB Introducing DuckLake

#ducklake #duckdb www.youtube.com/watch?v=zeon... I think if a ecosystem and more iceberg interoperability especially with different engines is available some day this will be an an awesome approach.

jack morris on X: "excited to finally share on arxiv what we've known for a while now: All Embedding Models Learn The Same Thing embeddings from different models are SO similar that we can map between them based on structure alone. without *any* paired data feels like magic, but it's real:🧵" / X excited to finally share on arxiv what we've known for a while now: All Embedding Models Learn The Same Thing embeddings from different models are SO similar that we can map between them based on structure alone. without *any* paired data feels like magic, but it's real:🧵

#ai #embeddings #privacy What are the implications of embedding #inversion ? x.com/jxmnop/statu...

YouTube video by Dominik Merli | Professor für IT-Sicherheit Cyber Gangsta’s Paradise | Prof. Merli ft. MC BlackHat [Parody Music Video]

#cyber #security #rap www.youtube.com/watch?v=6Hcs... #gangsta

GitHub - R9295/autarkie: Autarkie - Instant Grammar Fuzzing Using Rust Macros Autarkie - Instant Grammar Fuzzing Using Rust Macros - R9295/autarkie

cool #rust project - automatic grammar fuzzer github.com/R9295/autarkie

YouTube video by Black Hat SpAIware & More: Advanced Prompt Injection Exploits in LLM Applications

#blackhat www.youtube.com/watch?v=84NV...

Python sub-micro GPIO – PyXL Benchmark PyXL runs Python directly in hardware. GPIO toggle in sub-micro. See how it works.

#fpga #python executing python code directly from hardware www.runpyxl.com/gpio

www.ria.ee/en/state-inf...

GitHub - nordic-institute/X-Road: Source code of the X-Road® data exchange layer software Source code of the X-Road® data exchange layer software - nordic-institute/X-Road

#xroad #xtee truly impressive data exchange backbone github.com/nordic-insti...

interesting: The whole idea that developers should secure one part of the prompt from another part of the prompt is silly, and a strong signal that the abstractions we're using are broken. As this post makes clear: if any part of the prompt is in user space then the whole thing is in user space.

AI Horseless Carriages | koomen.dev An essay about bad AI app design

#horseless #carriages of #ai koomen.dev/essays/horse... do we use the right #abstractions?