4231
DeepSeek's Hidden Bias: How We Cut It by 76% Without Performance Loss
Showcasing Hirundo’s bias unlearning capabilities on DeepSeek-R1-Distill-Llama-8B, reducing its biases up to 76% compared to original performance.
New
The latest blog posts to take off on Hacker News. 31,672 posts.
4232
A major Postgres upgrade with zero downtime
Right before Christmas we discovered that our Aurora Postgres instance needed a major version upgrade. This covers how we made the upgrade with zero seconds of downtime.
4233
Case Study: ByteDance Uses eBPF to Enhance Networking Performance
Bytedance, a global technology company operating a wide range of content platforms around the world at massive scale, faced significant challenges in ensuring performance and stability across its data centers. With over a million servers running containerized applications, the company required a networking solution…
4234
Why DeepSeek had to be open source
DeepSeek trained a model as good as o1 at a fraction of the cost, then gave it away. Here's why that wasn't a mistake and what it means for OpenAI.
4235
The Debian Publicity Team will no longer post on X/Twitter
The Debian Publicity Team will no longer post on X/Twitter. We took this decision since we feel X doesn't reflect Debian shared values as stated in our social contract, code of conduct and diversity statement. X evolved into a place where people we care about don't feel safe. You are very much invited to follow us on…
4236
4237
Questions censored by DeepSeek
Analysis of DeepSeek-R1 censorship using 1,156 political prompts exposing CCP content filtering and bias detection patterns
4238
How has DeepSeek improved the Transformer architecture?
This Gradient Updates issue goes over the major changes that went into DeepSeek's most recent model.
4239
Qwen2.5-Max: Exploring the intelligence of large-scale MoE model
QWEN CHAT API DEMO DISCORD It is widely recognized that continuously scaling both data size and model size can lead to significant improvements in model intelligence. However, the research and industry community has limited experience in effectively scaling extremely large models, whether they are dense or…
4240
IAC confirms existence of a Super-earth in the habitable zone of a Sun-like Star
The Instituto de Astrofísica de Canarias (IAC) and the Universidad de La Laguna (ULL) has confirmed the discovery of a super-Earth orbiting in the habitable zone of HD 20794, a nearby Sun-like star. This discovery, the result of over two decades of observations, opens a window to future studies of Earth-like planetary…
4241
Run DeepSeek R1 Dynamic 1.58-bit
DeepSeek R-1 is the most powerful open-source reasoning model that performs on par with OpenAI's o1 model. Run the 1.58-bit Dynamic GGUF version by Unsloth.
4242
Why OpenAI's $157B valuation misreads AI's future (Oct 2024)
I break down the logic behind the company's towering price tag and why I think the most valuable AI companies have yet to be built.
4243
Google open-sources the Pebble OS
The source code for the Pebble operating system in now open, to support the volunteer community maintaining Pebble watches.
4244
The future of Rebble
Today we’re excited to announce several developments which will affect the future of Rebble. Let’s get straight into it, starting with the big one…🎉 Google O...
4245
Operation Leg – a pilot unlike any other (2020)
Douglas Bader was a Battle of Britain pilot unlike any other. Medically discharged against his will in 1932, the outbreak of the war was an opportunity for Bader to re-join the RAF and take back to the skies.
4246
4247
Render raises $80M in Series C financing
Today marks two significant milestones in Render's journey: we’ve surpassed 2 million developers on the platform and added $80M in Series C funding to reimagine the cloud for the AI era.
4248
No Bitcoin ETFs at Vanguard (2024)
Vanguard's thought leaders share timely insights about international financial markets, investor behavior, and the investment industry.
4249
Qwen2.5-1M: Deploy your own Qwen with context length up to 1M tokens
Tech Report HuggingFace ModelScope Qwen Chat HuggingFace Demo ModelScope Demo DISCORD Introduction Two months after upgrading Qwen2.5-Turbo to support context length up to one million tokens, we are back with the open-source Qwen2.5-1M models and the corresponding inference framework support. Here’s what you can…
4250
The protester's guide to smartphone security
Your phone is an essential tool, but it also represents a huge risk to your privacy and security. Understanding these best practices when it comes to securing your smartphone will help keep you and your data safe.
4251
Emerging reasoning with reinforcement learning
Made in Notion, the connected workspace where AI agents and teams organize docs, projects and knowledge together.
4252
Texas Is Enforcing Its State Data Privacy Law. So Should Other States
States need to have and use data privacy laws to bring privacy violations to light and hold companies accountable for them. So, we were glad to see that the Texas Attorney General’s Office has filed its first lawsuit under Texas Data Privacy and Security Act (TDPSA) to take the Allstate Corporation...
4253
Caltrain's electric fleet more efficient than expected
Caltrain announced at its monthly Board of Directors meeting that regenerative braking on the new trains is generating and sending back to the electric grid approximately 23% of the energy consumed by the system.
4254
The hidden complexity of scaling WebSockets
WebSockets are at the very core of Compose. Learn how we scaled them to support 1000s of concurrent users.
4255
4256
A WebAssembly compiler that fits in a tweet
Starting with a 192-byte one-liner that implements a Reverse Polish Notation arithmetic compiler, we'll work backward to transform it into readable JavaScript by removing one code golf trick at a time
4257
Threlte 8
The next major release of Threlte
4258
Every System is a Log: Avoiding coordination in distributed applications
Distributed coordination makes application complex and brittle. Because all systems eventually build on logs, we can use a shared-log approach to eliminate most coordination. We discuss this conceptually and show how this is practically implemented in the open source project Restate.
4259
4260
How we scaled Slack to support 1000s of developers
Here’s how we built a two-way Slack bridge with full impersonation to deliver white glove support to thousands of Railway developers from our help center.