331
Fuyu-8B: A multimodal architecture for AI agents
We’re open-sourcing Fuyu-8B - a small version of the multimodal model that powers our product.
Most cited
Posts that commenters keep linking to, ranked by how many different people have cited them in Hacker News comments. 363 posts.
332
Introduction to Linux interfaces for virtual networking (2018)
Explore common Linux virtual network interfaces, including Bridge, Bonded interface, VLAN, VXLAN, MACVLAN, IPVLAN, MACVTAP/IPVTAP, and more. Learn how to create and use them effectively for your
333
Ruby 3.3's YJIT: Faster While Using Less Memory
This year, the YJIT team has been working hard to improve and optimize YJIT. We’re proud to say that the version of YJIT to be included with Ruby 3.3 is leaps and bounds ahead of Ruby 3.2’s. It provides better performance across the board, while also warming up faster and using less memory. The 3.3 release is also…
334
Polars: Company Formation Announcement
We are excited to announce the start of company that will build around Polars, enabling data processing at any scale.
335
Removing PGP from PyPI
PyPI has removed support for uploading PGP signatures with new releases.
336
Swedish union declares strike for Tesla employees
The main reason for IF Metall to take industrial action at Tesla is to ensure that our members have decent and safe working conditions. Over a long period of time, we have attempted to discuss with Tesla the signing of a collective agreement, yet without success. Now we see no solution other than to take industrial…
337
Culture eats policy
The government has a culture problem, making it difficult to implement well-intentioned policies.
338
Ruby 3.3's YJIT Runs Shopify's Production Code 15% Faster
Ruby 3.2 YJIT has optimized the production workloads of Shopify and other companies. We encourage you to enable YJIT in production. Once Ruby 3.3 is released, it should make your application even faster.
339
WASI Support in Go
Go 1.21 adds a new port targeting the WASI preview 1 syscall API
340
S3 Express Is All You Need
The future of modern data infrastructure is object storage.
341
Improving mathematical reasoning with process supervision
We’ve trained a model to achieve a new state-of-the-art in mathematical problem solving by rewarding each correct step of reasoning (“process supervision”) instead of simply rewarding the correct final answer (“outcome supervision”). In addition to boosting performance relative to outcome supervision, process…
342
343
Dart 3.1 and a retrospective on functional style programming in Dart
Pattern matching and exhaustive switches come together to enable functional style data models that blend seamlessly with Dart’s…
344
No Google Topics in Vivaldi
Spying on people’s behavior & profiling them is wrong. That’s why we have made sure that Google’s Topics is disabled in two separate ways in Vivaldi browser.
345
Llama 2 is about as factually accurate as GPT-4 for summaries and is 30X cheaper
Using Anyscale Endpoints, we compared Llama 2 7b, 13b and 70b vs. OpenAI's GPT-3.5-turbo and GPT-4 for accuracy and cost.
346
347
Hardware Intrinsics in .NET 8
.NET 8 includes significant improvements to the Hardware Intrinsics feature.
348
NYPD can use more than 15,000 cameras to track people using facial recognition
Thousands of volunteers helped map surveillance cameras that the NYPD could use to track people using facial recognition.
349
How to use Podman inside of a container (2021)
One of the most asked about topics to folks working on upstream container technologies is running Podman within a container. Most of this has histo...
350
Diffusion with Offset Noise: Finetuning SD to generate very dark or light images
Historical research record. Fine-tuning against a modified noise, enables Stable Diffusion to generate very dark or light images easily.
351
This is what the year actually looks like (2018)
A few weeks back, I asked the Norwegian public what the year looks like inside their heads. It led to a deluge of responses. Here's a summary of more than 75 000 month placements, and nearly 40 000 questionnaire responses.
352
EU member states still cannot agree about end-to-end encryption
Stanford Center for Internet and Society2023119 points147 commentscited by 3Society, Policy & Culture details
Recently, Wired reported on documents leaked in May from the Council of the European Union’s ongoing deliberations over the draft Child Sex Abuse Regulation. The proposal is controversial because, as presently written, it would require online service providers to proactively scan their services – including users’…
353
DuckDuckGo Launches DuckAssist
A new feature that generates natural language answers to search queries using Wikipedia.
354
Protecting customers with generative AI indemnification
Google Cloud assumes responsibility for potential legal risks of using our generative AI, offering indemnities for training data and generated output.
355
Why is building a UI in Rust so hard?
Hear from a Warp engineer why building a UI in Rust is challenging. Learn about innovative solutions and approaches to tackle its strict memory management and lack of inheritance.
356
Continuous batching to increase LLM inference throughput and reduce p50 latency
In this blog, we discuss continuous batching, a critical systems-level optimization that improves both throughput and latency under load for LLMs.
357
User interface design timeframes: from 0.1 seconds to 10 years (2009)
From 0.1 seconds to 10 years or more, user interface design has many different timeframes, and each has its own particular usability issues.
358
Next.js 14
Next.js 14 includes included performance, stability for Server Actions, a new course teaching the App Router, and more.
359
What is gained and lost with 63-bit integers? (2014)
Almost every programming language uses 64-bit integers on typical modern Intel machines. OCaml uses a special 63-bit representation. How does it affect OCaml?
360
Apache Hudi vs. Delta Lake vs. Apache Iceberg Lakehouse Feature Comparison
A thorough comparison of the Apache Hudi™, Delta Lake, and Apache Iceberg™ data lakehouse projects across features, community, and performance benchmarks. This includes a focus on common use cases such as change data capture (CDC) and data ingestion.