By Daisuke Wakabayashi from NYT Business https://ift.tt/e8qoWlT
Chinese Factories Slow in Early Sign of Trade War’s Toll
By Daisuke Wakabayashi from NYT Business https://ift.tt/e8qoWlT
Word of the Day: runic
By The Learning Network from NYT The Learning Network https://ift.tt/g4deEXF
As India and Pakistan Cancel Visas, Parents and Children Are Separated
By Suhasini Raj, Mujib Mashal, Pragati K.B. and Atul Loke from NYT World https://ift.tt/oPgdtyT
Spelling Bee Forum
By New York Times Games from NYT Gameplay https://ift.tt/7EjXaiA
New top story on Hacker News: Show HN: Remote-Controlled IKEA Deathstar Lamp
Show HN: Remote-Controlled IKEA Deathstar Lamp
8 by sephalon | 0 comments on Hacker News.
Repainting the iconic IKEA PS 2014 lamp into the Deathstar from Star Wars has been a popular IKEA hack for quite some time. This variant additionally replaces the manual, rope-operated mechanism to open and close the lamp with a remote-controlled motor. The firmware is based on ESPHome, and its excellent Home Assistant integration enables one to implement higher-level features, like a "sundial" where the aperture of the Deathstar follows the sun elevation throughout the day (see the timelapse video). That said, I will not consider this project as complete until the Imperial March can be played over the stepper motor (just like the legendary Floppotron) ;-)
8 by sephalon | 0 comments on Hacker News.
Repainting the iconic IKEA PS 2014 lamp into the Deathstar from Star Wars has been a popular IKEA hack for quite some time. This variant additionally replaces the manual, rope-operated mechanism to open and close the lamp with a remote-controlled motor. The firmware is based on ESPHome, and its excellent Home Assistant integration enables one to implement higher-level features, like a "sundial" where the aperture of the Deathstar follows the sun elevation throughout the day (see the timelapse video). That said, I will not consider this project as complete until the Imperial March can be played over the stepper motor (just like the legendary Floppotron) ;-)
New top story on Hacker News: Show HN: Faasta – A self-hosted Serverless platform for WASM-wasi-HTTP in Rust
Show HN: Faasta – A self-hosted Serverless platform for WASM-wasi-HTTP in Rust
19 by alexboehm | 8 comments on Hacker News.
I've just released an early version off my project I've been working on for a few months now and would love some feedback. https://ift.tt/j6lbnLJ I was surprised there isn't yet an open/source and standards compliant way to host wasi-http functions in a way that takes advantages of WASM, a multi tenanted application. If you're not familiar with wasi, Compared to something like AWS Lambda, this approach is much more efficient as a single process can serve 1000s of function invocations concurrently and asynchronously, instead requiring an entire VM. This is still early days for the project, but feel free to download the cli utility with cargo install cargo-faasta. Feel free to test deploying functions on my hosted instance at https://website.faasta.xyz . The service is free to use and currently supports deployments via GitHub OAuth, with a limit of 10 functions per GitHub account.
19 by alexboehm | 8 comments on Hacker News.
I've just released an early version off my project I've been working on for a few months now and would love some feedback. https://ift.tt/j6lbnLJ I was surprised there isn't yet an open/source and standards compliant way to host wasi-http functions in a way that takes advantages of WASM, a multi tenanted application. If you're not familiar with wasi, Compared to something like AWS Lambda, this approach is much more efficient as a single process can serve 1000s of function invocations concurrently and asynchronously, instead requiring an entire VM. This is still early days for the project, but feel free to download the cli utility with cargo install cargo-faasta. Feel free to test deploying functions on my hosted instance at https://website.faasta.xyz . The service is free to use and currently supports deployments via GitHub OAuth, with a limit of 10 functions per GitHub account.
Word of the Day: simulacrum
By The Learning Network from NYT The Learning Network https://ift.tt/r91xnah
At Least 9 Dead in Russian Airstrikes on Kyiv, Ukrainian Officials Say
By Kim Barker and Maria Varenikova from NYT World https://ift.tt/3WgoBms
Word of the Day: tetchy
By The Learning Network from NYT The Learning Network https://ift.tt/NdWl5Xx
Word of the Day: reverberation
By The Learning Network from NYT The Learning Network https://ift.tt/DAE3vqW
Hidden Above a Trap Door, 17th-Century Frescoes Come to Light
By Franz Lidz from NYT Science https://ift.tt/q89lmUn
Word of the Day: quasi
By The Learning Network from NYT The Learning Network https://ift.tt/MN5QvmD
Spelling Bee Forum
By New York Times Games from NYT Gameplay https://ift.tt/9IZktGR
Spelling Bee Forum
By New York Times Games from NYT Gameplay https://ift.tt/OIZwvf4
New top story on Hacker News: Show HN: New world record – verified Goldbach Conjecture up to 4*10^18+7*10^13
Show HN: New world record – verified Goldbach Conjecture up to 4*10^18+7*10^13
40 by jay_gridbach | 3 comments on Hacker News.
Achieved a new world record in verifying the Goldbach Conjecture using grid computing, by extending the verification up to 4 quadrillion (4×10¹⁸) + 70 trillion (7×10¹³). My grid computing system - Gridbach is a cloud-based distributed computing system accessible from any PC or smartphone. It requires no login or app installation. The high-performance WASM (WebAssembly) binary code is downloaded as browser content, enabling computation on the user’s browser. [Website] https://gridbach.com/ [Medium] https://ift.tt/gDlpkyX...
40 by jay_gridbach | 3 comments on Hacker News.
Achieved a new world record in verifying the Goldbach Conjecture using grid computing, by extending the verification up to 4 quadrillion (4×10¹⁸) + 70 trillion (7×10¹³). My grid computing system - Gridbach is a cloud-based distributed computing system accessible from any PC or smartphone. It requires no login or app installation. The high-performance WASM (WebAssembly) binary code is downloaded as browser content, enabling computation on the user’s browser. [Website] https://gridbach.com/ [Medium] https://ift.tt/gDlpkyX...
Tomó un siglo encontrar este calamar colosal
By Elizabeth Preston from NYT En español https://ift.tt/jKia34e
Word of the Day: chortle
By The Learning Network from NYT The Learning Network https://ift.tt/toxGne5
Stephen Colbert Defends PBS and NPR Against Trump’s Defunding Plan
By Trish Bendix from NYT Arts https://ift.tt/PcdFiVJ
Spelling Bee Forum
By New York Times Games from NYT Gameplay https://ift.tt/a0kPLoV
New top story on Hacker News: Show HN: I Made Bagel Font
Show HN: I Made Bagel Font
14 by hmpark | 10 comments on Hacker News.
Hey HN, I recently made 3d style Bagel web font (with some more styles) from scratch, and wanted to share the process in case anyone's interested in experimenting or giving feedback. You can play around by typing https://ift.tt/wk3bXzS [Backstory] I was experimenting with Stable Diffusion + ControlNet to generate icons. To my surprise, generating letters this way was actually better than using traditional 3D tools like Blender for this purpose. That led to the idea: what if I could turn into an actual real working font? I found a few tools, but most were either paid or lacked support. So I leaned on Cursor (been using it since launch) and spent a few hours hacking together a Python script using FontTools( https://ift.tt/hubdOa3 , Opensource font generation python library). Eventually, I got a working .ttf file with an SBIX table (used by Apple’s color fonts)! [Technical details] 1) Create base glyphs Drew black letterforms by hand on iPad (or used open-licensed fonts for other experiments). 2) Add texture/style Used ComfyUI + Stable Diffusion + ControlNet to generate stylized images for each letter. Could also be done in Midjourney, Photoshop, etc. Key part is keeping image sizes consistent. 3) Generate font files Used FontTools (Python). Tried FontForge, BitFontMaker, even paid tools like Glyphs and Fontself — none worked well for image-based fonts. FontTools + SBIX table finally did the trick. [Constraints] SBIX fonts are still relatively heavy, and not all software supports them yet. Photoshop does, but Figma doesn’t (yet). I plan to submit a feature request to Figma once things are more polished. If anyone’s curious to try the font, happy to share it. Would love feedback, ideas, or just to know if this inspires you. I’ve been doing solo creative projects for a while — it can get tough, but knowing it sparks something for someone else helps a lot. Thanks for reading — and feel free to AMA if you're trying something similar! *Updates I'm using a Mac with the Chrome (Arc) browser. It might not work properly on other browsers or OS. If it doesn't work for you, reporting it would really help me improve it next time. Thanks!!
14 by hmpark | 10 comments on Hacker News.
Hey HN, I recently made 3d style Bagel web font (with some more styles) from scratch, and wanted to share the process in case anyone's interested in experimenting or giving feedback. You can play around by typing https://ift.tt/wk3bXzS [Backstory] I was experimenting with Stable Diffusion + ControlNet to generate icons. To my surprise, generating letters this way was actually better than using traditional 3D tools like Blender for this purpose. That led to the idea: what if I could turn into an actual real working font? I found a few tools, but most were either paid or lacked support. So I leaned on Cursor (been using it since launch) and spent a few hours hacking together a Python script using FontTools( https://ift.tt/hubdOa3 , Opensource font generation python library). Eventually, I got a working .ttf file with an SBIX table (used by Apple’s color fonts)! [Technical details] 1) Create base glyphs Drew black letterforms by hand on iPad (or used open-licensed fonts for other experiments). 2) Add texture/style Used ComfyUI + Stable Diffusion + ControlNet to generate stylized images for each letter. Could also be done in Midjourney, Photoshop, etc. Key part is keeping image sizes consistent. 3) Generate font files Used FontTools (Python). Tried FontForge, BitFontMaker, even paid tools like Glyphs and Fontself — none worked well for image-based fonts. FontTools + SBIX table finally did the trick. [Constraints] SBIX fonts are still relatively heavy, and not all software supports them yet. Photoshop does, but Figma doesn’t (yet). I plan to submit a feature request to Figma once things are more polished. If anyone’s curious to try the font, happy to share it. Would love feedback, ideas, or just to know if this inspires you. I’ve been doing solo creative projects for a while — it can get tough, but knowing it sparks something for someone else helps a lot. Thanks for reading — and feel free to AMA if you're trying something similar! *Updates I'm using a Mac with the Chrome (Arc) browser. It might not work properly on other browsers or OS. If it doesn't work for you, reporting it would really help me improve it next time. Thanks!!
Word of the Day: soporific
By The Learning Network from NYT The Learning Network https://ift.tt/YVO9zlF
Spelling Bee Forum
By New York Times Games from NYT Gameplay https://ift.tt/LBjQOZy
Word of the Day: invective
By The Learning Network from NYT The Learning Network https://ift.tt/h9UMd0o
Word of the Day: vermilion
By The Learning Network from NYT The Learning Network https://ift.tt/Psham2W
Spelling Bee Forum
By New York Times Games from NYT Gameplay https://ift.tt/hWz325N
New top story on Hacker News: Show HN: Chonky – a neural approach for text semantic chunking
Show HN: Chonky – a neural approach for text semantic chunking
12 by hessdalenlight | 0 comments on Hacker News.
TLDR: I’ve made a transformer model and a wrapper library that segments text into meaningful semantic chunks. The current text splitting approaches rely on heuristics (although one can use neural embedder to group semantically related sentences). I propose a fully neural approach to semantic chunking. I took the base distilbert model and trained it on a bookcorpus to split concatenated text paragraphs into original paragraphs. Basically it’s a token classification task. Model fine-tuning took day and a half on a 2x1080ti. The library could be used as a text splitter module in a RAG system or for splitting transcripts for example. The usage pattern that I see is the following: strip all the markup tags to produce pure text and feed this text into the model. The problem is that although in theory this should improve overall RAG pipeline performance I didn’t manage to measure it properly. Other limitations: the model only supports English for now and the output text is downcased. Please give it a try. I'll appreciate a feedback. The Python library: https://ift.tt/4JC1Lq5 The transformer model: https://ift.tt/lnMSaVd...
12 by hessdalenlight | 0 comments on Hacker News.
TLDR: I’ve made a transformer model and a wrapper library that segments text into meaningful semantic chunks. The current text splitting approaches rely on heuristics (although one can use neural embedder to group semantically related sentences). I propose a fully neural approach to semantic chunking. I took the base distilbert model and trained it on a bookcorpus to split concatenated text paragraphs into original paragraphs. Basically it’s a token classification task. Model fine-tuning took day and a half on a 2x1080ti. The library could be used as a text splitter module in a RAG system or for splitting transcripts for example. The usage pattern that I see is the following: strip all the markup tags to produce pure text and feed this text into the model. The problem is that although in theory this should improve overall RAG pipeline performance I didn’t manage to measure it properly. Other limitations: the model only supports English for now and the output text is downcased. Please give it a try. I'll appreciate a feedback. The Python library: https://ift.tt/4JC1Lq5 The transformer model: https://ift.tt/lnMSaVd...
How an American Sign Language Artist Spends Their Sundays
By Sarah Bahr from NYT New York https://ift.tt/GJCEMwc
Spelling Bee Forum
By New York Times Games from NYT Gameplay https://ift.tt/sYc71Um
Senate Confirms New Chairman of Joint Chiefs of Staff
By Helene Cooper from NYT U.S. https://ift.tt/ZB2TLHY
Word of the Day: dormant
By The Learning Network from NYT En español https://ift.tt/dbzm3K5
Reeling From Trump’s Tariffs, South Korea Tries Striking a Delicate Balance
By Choe Sang-Hun from NYT World https://ift.tt/vEm5p8N
Global Art Sales Fell by 12% Last Year, Report Says
By Scott Reyburn from NYT Arts https://ift.tt/a3AbNxk
Word of the Day: artisanal
By The Learning Network from NYT The Learning Network https://ift.tt/oIjX6uz
China Tries to Downplay the Trade War’s Effects on Its Economy
By David Pierson, Claire Fu and Vivian Wang from NYT World https://ift.tt/ofivjhC
Spelling Bee Lets Solvers Have Plurals (for One Day)
By Lyna Bentahar from NYT Corrections https://ift.tt/yzl3tsJ
Spelling Bee Forum
By New York Times Games from NYT Gameplay https://ift.tt/hT4Z5or
Stephen Colbert Likens Trump’s Tariffs to an Economic Infection
By Trish Bendix from NYT Arts https://ift.tt/7gTwM5y
Word of the Day: flaxen
By The Learning Network from NYT The Learning Network https://ift.tt/Ql35oD8
Word of the Day: rapprochement
By The Learning Network from NYT The Learning Network https://ift.tt/VKJY8v3
New top story on Hacker News: Show HN: Duolingo-style exercises but with real-world content like the news
Show HN: Duolingo-style exercises but with real-world content like the news
22 by ph4evers | 6 comments on Hacker News.
I've been working on a little side project that combines Duolingo-like listening comprehension exercises with real content . Every video is transcribed to get much better transcripts than the closed captions. I filter on high quality transcripts, and afterwards a LLM selects only plausible segments for the exercises. This seems to work well for quality control and seems to be reliable enough for these short exercises. Would love your thoughts!
22 by ph4evers | 6 comments on Hacker News.
I've been working on a little side project that combines Duolingo-like listening comprehension exercises with real content . Every video is transcribed to get much better transcripts than the closed captions. I filter on high quality transcripts, and afterwards a LLM selects only plausible segments for the exercises. This seems to work well for quality control and seems to be reliable enough for these short exercises. Would love your thoughts!
China Holds Military Exercises Around Taiwan in Fresh Warning
By Chris Buckley from NYT World https://ift.tt/k2zRnHy