Notes on OpenAI's new o1 chain-of-thought models (🔥 Score: 158+ in 3 hours)
Link: https://readhacker.news/s/6eu69
Comments: https://readhacker.news/c/6eu69
Ergo: Erlang-inspired event driven actor framework in Go (Score: 150+ in 15 hours)
Link: https://readhacker.news/s/6erD9
Comments: https://readhacker.news/c/6erD9
1913: When Hitler, Trotsky, Tito, Freud and Stalin all lived in the same place (Score: 156+ in 7 hours)
Link: https://readhacker.news/s/6espa
Comments: https://readhacker.news/c/6espa
OpenAI O1 Model (🔥 Score: 168+ in 21 minutes)
Link: https://readhacker.news/s/6esMq
Comments: https://readhacker.news/c/6esMq
Show HN: iFixit created a new USB-C, repairable soldering system (🔥 Score: 165+ in 1 hour)
Link: https://readhacker.news/s/6esqR
Comments: https://readhacker.news/c/6esqR
After years of making screwdrivers and teaching people to repair electronics, we just made our first electronic tool. It's been a journey for us to build while hewing to our repairable principles. We're really excited about it.
It's a USB-C powered soldering iron and smart battery power hub. Super repairable, of course. Our goal is to make soldering so easy everyone can do it:
https://www.ifixit.com/fixhub
We didn’t want to make just another iron, so we spent years sweating the details and crafting something that met our exacting standards. This is a high-performance iron: it can output 100W of heat, gets to soldering temperature in under 5 seconds, and automatically cools off when you set it down. The accelerometer detects when you pick it up and heats it back up. Keeping the iron at a lower temperature while you’re not soldering shouold prolong the life of the tip.
What’s the difference between this iron and other USB-C irons on the market? Here’s a quick list:
Higher power (our Smart Iron is 100W, competitors max out at 60W over USB-C, 88W over DC Supply)
Heat-resistant storage cap (you just have to try this out, it’s a real game changer in day-to-day use)
Polished user experience
A warranty and a local company to talk to (I can’t find any contact information for Miniware)
Comfier / more natural grip
Shorter soldering tip length
No-tangle, heat-resistant cable
Locking ring on the cable, so it can’t snag and get disconnected (this happens to me all the time on other irons)
More intuitive settings, either on the Power Station or on the computer
We used [Web Serial](https://caniuse.com/web-serial) for the interface, which is only supported in Chromium browsers. The biggest bummer with that is that no mobile browsers support it, yet. Hopefully that changes soon.
Hardware is hard! It's been a journey for us. Happy to answer any questions about how we made it.
Schematics and repair information are online here: https://www.ifixit.com/Device/FixHub_Portable_Soldering_Stat...
Reader-LM: Small Language Models for Cleaning and Converting HTML to Markdown (Score: 153+ in 17 hours)
Link: https://readhacker.news/s/6eqsk
Comments: https://readhacker.news/c/6eqsk
SpaceX Astronauts Begin Spacewalk, Putting New Spacesuits to Test (🔥 Score: 150+ in 3 hours)
Link: https://readhacker.news/s/6erFR
Comments: https://readhacker.news/c/6erFR
Going open-source as a VC-Backed company (Score: 150+ in 1 day)
Link: https://readhacker.news/s/6ejAu
Comments: https://readhacker.news/c/6ejAu
A Uruguayan company teaches people how to turn regular cars into EVs (Score: 150+ in 19 hours)
Link: https://readhacker.news/s/6epvy
Comments: https://readhacker.news/c/6epvy
Show HN: Tune LLaMa3.1 on Google Cloud TPUs (Score: 150+ in 19 hours)
Link: https://readhacker.news/s/6epig
Comments: https://readhacker.news/c/6epig
Hey HN, we wanted to share our repo where we fine-tuned Llama 3.1 on Google TPUs. We’re building AI infra to fine-tune and serve LLMs on non-NVIDIA GPUs (TPUs, Trainium, AMD GPUs).
The problem: Right now, 90% of LLM workloads run on NVIDIA GPUs, but there are equally powerful and more cost-effective alternatives out there. For example, training and serving Llama 3.1 on Google TPUs is about 30% cheaper than NVIDIA GPUs.
But developer tooling for non-NVIDIA chipsets is lacking. We felt this pain ourselves. We initially tried using PyTorch XLA to train Llama 3.1 on TPUs, but it was rough: xla integration with pytorch is clunky, missing libraries (bitsandbytes didn't work), and cryptic HuggingFace errors.
We then took a different route and translated Llama 3.1 from PyTorch to JAX. Now, it’s running smoothly on TPUs! We still have challenges ahead, there is no good LoRA library in JAX, but this feels like the right path forward.
Here's a demo (https://dub.sh/felafax-demo) of our managed solution.
Would love your thoughts on our repo and vision as we keep chugging along!
Be a Thermostat, Not a Thermometer (Score: 151+ in 8 hours)
Link: https://readhacker.news/s/6eqCZ
Comments: https://readhacker.news/c/6eqCZ
Max Headroom and the World of Pseudo-CGI (2013) (Score: 150+ in 22 hours)
Link: https://readhacker.news/s/6enu8
Comments: https://readhacker.news/c/6enu8
Scientists use magnetic nanotech to safely rewarm frozen tissues for transplant (Score: 150+ in 16 hours)
Link: https://readhacker.news/s/6epes
Comments: https://readhacker.news/c/6epes
The Golden Road: How Ancient India Transformed the World (Score: 150+ in 1 day)
Link: https://readhacker.news/s/6en4X
Comments: https://readhacker.news/c/6en4X
Transparenttextures.com (❄️ Score: 151+ in 2 days)
Link: https://readhacker.news/s/6eghv
Comments: https://readhacker.news/c/6eghv
GAZEploit: Remote keystroke inference attack by gaze estimation in VR/MR devices (Score: 150+ in 13 hours)
Link: https://readhacker.news/s/6erXN
Comments: https://readhacker.news/c/6erXN
Data sleuths who spotted research misconduct cleared of defamation (🔥 Score: 156+ in 3 hours)
Link: https://readhacker.news/s/6etDL
Comments: https://readhacker.news/c/6etDL
Metformin decelerates aging clock in male monkeys (Score: 150+ in 5 hours)
Link: https://readhacker.news/s/6esJV
Comments: https://readhacker.news/c/6esJV
DEF Con 32 – AMD Sinkclose Universal Ring-2 Privilege Escalation (Not Redacted) [pdf] (❄️ Score: 151+ in 4 days)
Link: https://readhacker.news/s/6ebLr
Comments: https://readhacker.news/c/6ebLr
Kolmogorov-Arnold networks may make neural networks more understandable (Score: 151+ in 5 hours)
Link: https://readhacker.news/s/6erz2
Comments: https://readhacker.news/c/6erz2
Why Oxide Chose Illumos (Score: 150+ in 18 hours)
Link: https://readhacker.news/s/6eqmh
Comments: https://readhacker.news/c/6eqmh
Why Haskell? (Score: 151+ in 6 hours)
Link: https://readhacker.news/s/6ermA
Comments: https://readhacker.news/c/6ermA
NASA Pulls Off Delicate Thruster Swap, Keeping Voyager 1 Mission Alive (Score: 150+ in 9 hours)
Link: https://readhacker.news/s/6eqVS
Comments: https://readhacker.news/c/6eqVS
Large-Scale Generation of Transit Maps from OpenStreetMap Data (❄️ Score: 151+ in 2 days)
Link: https://readhacker.news/s/6efQa
Comments: https://readhacker.news/c/6efQa
Show HN: Konty – A Balsamiq-alternative lo-fi wireframe tool for modern apps (Score: 154+ in 4 hours)
Link: https://readhacker.news/s/6eqWA
Comments: https://readhacker.news/c/6eqWA
How we made Jupyter notebooks load faster (Score: 150+ in 1 day)
Link: https://readhacker.news/s/6ejAL
Comments: https://readhacker.news/c/6ejAL
Feeld dating app – Your nudes and data were publicly available (Score: 154+ in 10 hours)
Link: https://readhacker.news/s/6eqnH
Comments: https://readhacker.news/c/6eqnH
Mistral releases Pixtral 12B, its first multimodal model (Score: 151+ in 9 hours)
Link: https://readhacker.news/s/6eq8r
Comments: https://readhacker.news/c/6eq8r
My business card runs Linux and Ultrix (2022) (Score: 150+ in 4 hours)
Link: https://readhacker.news/s/6eqFE
Comments: https://readhacker.news/c/6eqFE
Noisy neighbor detection with eBPF (Score: 152+ in 9 hours)
Link: https://readhacker.news/s/6epQW
Comments: https://readhacker.news/c/6epQW