Hebrish: English-Hebrew code-switching
Collection
Every Hebrish model, dataset and Space, including those still on danielrosehill. hebrish.app • 34 items • Updated
id string | text string | target_terms list | categories list | homophone_set null | split string | derivation string | sentence_author string | source_dataset string | source_licence string | source_url string | source_authors list | source_date timestamp[s] | original_text string | review_note string | term_origins list |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
p_0009789919 | My home server is a Ryzen AI Max+ 395 system running Linux Fedora 43 with KDE and Wayland. | [
"Ryzen",
"Linux",
"Fedora",
"KDE",
"Wayland"
] | [
"tech",
"numeric"
] | null | train | mined-verbatim | pszemraj/LocalLLaMA-posts | pszemraj/LocalLLaMA-posts | ODC-BY-1.0 | https://www.reddit.com/r/LocalLLaMA/comments/1rjh5wg/unsloth_fixed_version_of_qwen3535ba3b_is/ | [] | 2026-03-03T00:00:00 | null | null | [
"claude",
"daniel",
"github-topics",
"daniel",
"claude"
] |
p_000a0d74f7 | Reminder to myself: try the new Hunyuan video model this weekend and compare the results against what I got from the other open models. | [
"Hunyuan"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"openrouter"
] |
p_0064a29e12 | Please run the install script with sudo, but first check that it does not overwrite the existing config on the server. | [
"sudo"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"daniel"
] |
p_007904719b | Look through the docs folder and summarize which papers from DeepMind we reference, then add a short section to the readme listing them with one line each. | [
"DeepMind"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_00da303bff | Set up a test run with Devstral on this repo and see whether it can fix the failing tests without touching the config. | [
"Devstral"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"openrouter"
] |
p_00e0e26a15 | Our 1B model achieves exceptional performance, surpassing GPT-3.5-Turbo and Claude 3 Haiku. | [
"GPT",
"Claude",
"Haiku"
] | [
"tech",
"numeric"
] | null | train | mined-edited | hysts-bot-data/daily-papers | hysts-bot-data/daily-papers | CC0-1.0 (compilation; arXiv metadata, including abstracts, is CC0) | https://arxiv.org/abs/2406.18518 | [
"Zuxin Liu",
"Thai Hoang",
"Jianguo Zhang",
"Ming Zhu",
"Tian Lan",
"Shirley Kokane",
"Juntao Tan",
"Weiran Yao",
"Zhiwei Liu",
"Yihao Feng",
"Rithesh Murthy",
"Liangwei Yang",
"Silvio Savarese",
"Juan Carlos Niebles",
"Huan Wang",
"Shelby Heinecke",
"Caiming Xiong"
] | 2024-06-26T00:00:00 | Moreover, our 1B model achieves exceptional performance, surpassing GPT-3.5-Turbo and Claude-3 Haiku. | fixed: stripped leading 'Moreover,'; Claude-3 Haiku -> Claude 3 Haiku | [
"daniel",
"daniel",
"claude"
] |
p_00e8242a73 | For our production architecture, we're considering a hybrid approach using GPT-4o for complex reasoning and Mixtral for general text generation tasks. | [
"GPT-4o",
"Mixtral"
] | [
"tech"
] | null | train | borrowed | danielrosehill/Tech-Sentences-For-ASR-Training | danielrosehill/Tech-Sentences-For-ASR-Training | MIT | https://huggingface.co/datasets/danielrosehill/Tech-Sentences-For-ASR-Training | [] | null | null | null | [
"daniel",
"openrouter"
] |
p_01033a13e0 | You can also edit images directly on your screen using the Nano Banana model. | [
"Nano Banana"
] | [
"tech"
] | null | train | mined-verbatim | e1879/threads-english-tech-news-sft | e1879/threads-english-tech-news-sft | CC-BY-4.0 | https://www.threads.net/@daily_ai_smart_tools/post/DUHQbJsiGcC | [] | 2026-01-29T00:00:00 | null | null | [
"daniel"
] |
p_015751cbc8 | Note to self: the weird lag on the smart home is probably dnsmasq handing out the wrong DNS server, so check the config tonight. | [
"dnsmasq"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_015c5af949 | Can you take a look at the OpenCV script that crops the camera frames? It's running slowly on the Raspberry Pi and I think the resize step is the problem. | [
"OpenCV"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"github-topics"
] |
p_01894d87af | Hi Sarah, I switched the whole office router to NextDNS last night, so if any site looks blocked, let me know and I'll whitelist it. | [
"NextDNS"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_018eec6167 | Please look through the repo and write a small script that spins up a fresh DigitalOcean droplet, installs Docker, and copies over my environment settings. Don't run anything yet, just show me the plan. | [
"DigitalOcean"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_018fc31b95 | I installed Termux on my old Android phone so I can SSH into the server and run a few scripts when I'm away from my laptop. | [
"Termux"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"github-topics"
] |
p_01ad6df5c3 | Hi Dana, I moved the staging cluster over to Argo CD last week, so deployments now follow whatever is merged in the repo. Let me know if anything looks off. | [
"Argo CD"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_01ae10142e | First, we'll create an RDS instance, then configure our VPC security groups to allow database connections from our application servers. | [
"RDS",
"VPC"
] | [
"tech"
] | null | train | borrowed | danielrosehill/Tech-Sentences-For-ASR-Training | danielrosehill/Tech-Sentences-For-ASR-Training | MIT | https://huggingface.co/datasets/danielrosehill/Tech-Sentences-For-ASR-Training | [] | null | null | null | [
"daniel",
"daniel"
] |
p_01dda853fd | I built it for OpenClaw, but it works with Codex, Cursor, Claude Code, or anything that speaks the OpenAI API. | [
"OpenClaw",
"Codex",
"Cursor",
"Claude Code",
"OpenAI"
] | [
"tech"
] | null | train | mined-verbatim | pszemraj/LocalLLaMA-posts | pszemraj/LocalLLaMA-posts | ODC-BY-1.0 | https://www.reddit.com/r/LocalLLaMA/comments/1r6dii2/nadirclaw_opensource_llm_router_for_openclaw_that/ | [] | 2026-02-16T00:00:00 | null | null | [
"daniel",
"daniel",
"claude",
"daniel",
"daniel"
] |
p_01deb2ff71 | Add a Grafana dashboard that shows CPU temperature and disk usage for the home server, and provision it from the repo so it survives a rebuild. | [
"Grafana"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_025d81d1d3 | Please look through the repo and tell me which scripts still call OpenCode directly, then replace them with a single wrapper so I can switch agents later. | [
"OpenCode"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_02a1f456c3 | To address practical challenges in large-scale deployment under diverse resource constraints, ERNIE 5.0 adopts a novel elastic training paradigm. | [
"ERNIE"
] | [
"tech",
"numeric"
] | null | train | mined-verbatim | hysts-bot-data/daily-papers | hysts-bot-data/daily-papers | CC0-1.0 (compilation; arXiv metadata, including abstracts, is CC0) | https://arxiv.org/abs/2602.04705 | [
"Haifeng Wang",
"Hua Wu",
"Tian Wu",
"Yu Sun",
"Jing Liu",
"Dianhai Yu",
"Yanjun Ma",
"Jingzhou He",
"Zhongjun He",
"Dou Hong",
"Qiwen Liu",
"Shuohuan Wang",
"Junyuan Shang",
"Zhenyu Zhang",
"Yuchen Ding",
"Jinle Zeng",
"Jiabin Yang",
"Liang Shen",
"Ruibiao Chen",
"Weichong Yin... | 2026-02-05T00:00:00 | null | null | [
"claude"
] |
p_02e3a354b5 | I'm using Lhotse to prepare the audio manifests for the NeMo training run. Can you check that the cuts look right before I start? | [
"Lhotse"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0310e82fdc | Reminder to myself to rewrite the staging setup in Pulumi this weekend, because clicking through the cloud console every time is driving me crazy. | [
"Pulumi"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_033eb6ce52 | Can you open TensorBoard for the latest run and check whether the validation loss is still going down, or if it flattened out? | [
"TensorBoard"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0352c6227a | Got Qwen Coder Next running on ROCm on my Strix Halo! | [
"Qwen",
"ROCm",
"Strix Halo"
] | [
"tech"
] | null | train | mined-edited | pszemraj/LocalLLaMA-posts | pszemraj/LocalLLaMA-posts | ODC-BY-1.0 | https://www.reddit.com/r/LocalLLaMA/comments/1qv65ed/got_qwencodernext_running_on_rocm_on_my_strix_halo/ | [] | 2026-02-03T00:00:00 | Got Qwen-Coder-Next running on ROCm on my Strix Halo! | fixed: Qwen-Coder-Next -> Qwen Coder Next | [
"daniel",
"claude",
"claude"
] |
p_0361c094c1 | Quick update on the server situation. The WhatsApp bot stopped replying last night because uvicorn crashed after I updated Ubuntu. I restarted it, and then I noticed the WAN link was dropping every few minutes. I think the router is the problem, not the app. I'll swap the cable tomorrow morning and watch the logs for a... | [
"WhatsApp",
"WAN",
"uvicorn",
"Ubuntu"
] | [
"long",
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"daniel",
"daniel",
"daniel",
"daniel"
] |
p_03a6b6b136 | Experiments across dense and Mixture-of-Experts LLMs demonstrate consistent gains on challenging reasoning benchmarks, including GPQA-Diamond, Omni-MATH, and HLE, with zero memory overhead and less than 2% latency increase. | [
"LLMs",
"GPQA",
"HLE"
] | [
"tech",
"numeric"
] | null | train | mined-verbatim | hysts-bot-data/daily-papers | hysts-bot-data/daily-papers | CC0-1.0 (compilation; arXiv metadata, including abstracts, is CC0) | https://arxiv.org/abs/2606.21906 | [
"Xuanming Zhang",
"Sining Zhoubian",
"Yuxuan Chen",
"Tianyi Tang",
"An Yang",
"Sean Du",
"Chujie Zheng",
"Fei Huang",
"Dayiheng Liu",
"Gao Huang",
"Jingren Zhou"
] | 2026-06-23T00:00:00 | null | null | [
"daniel",
"claude",
"claude"
] |
p_03ce2712bb | I gave Qwen3.5 35B a voice, a Telegram brain with over 25 tools, and remote access from my phone, all running on a Mac Studio M1 Ultra, zero cloud. | [
"Qwen3.5",
"Telegram"
] | [
"tech",
"numeric"
] | null | train | mined-edited | pszemraj/LocalLLaMA-posts | pszemraj/LocalLLaMA-posts | ODC-BY-1.0 | https://www.reddit.com/r/LocalLLaMA/comments/1rk6afk/free_code_realtime_voicetovoice_with_your_llm/ | [] | 2026-03-04T00:00:00 | I gave Qwen 3.5 35B a voice, a Telegram brain with 25+ tools, and remote access from my phone — all running on a Mac Studio M1 Ultra, zero cloud. | fixed: Qwen3.5, '25+' | [
"claude",
"daniel"
] |
p_03f3f6ab99 | We need to implement OAuth2 with OIDC to replace our legacy authentication system and enable proper single sign-on across all microservices. | [
"OAuth2",
"OIDC",
"single sign-on"
] | [
"tech"
] | null | train | borrowed | danielrosehill/Tech-Sentences-For-ASR-Training | danielrosehill/Tech-Sentences-For-ASR-Training | MIT | https://huggingface.co/datasets/danielrosehill/Tech-Sentences-For-ASR-Training | [] | null | null | null | [
"daniel",
"claude",
"daniel"
] |
p_0448a7d543 | We use this benchmark to evaluate nine LLM agents implemented using three different frameworks: aider, Claude Code, and OpenHands. | [
"LLM",
"Claude Code",
"OpenHands"
] | [
"tech"
] | null | train | mined-verbatim | hysts-bot-data/daily-papers | hysts-bot-data/daily-papers | CC0-1.0 (compilation; arXiv metadata, including abstracts, is CC0) | https://arxiv.org/abs/2506.22598 | [
"Nicholas Edwards",
"Yukyung Lee",
"Yujun",
"Mao",
"Yulu Qin",
"Sebastian Schuster",
"Najoung Kim"
] | 2025-07-01T00:00:00 | null | null | [
"daniel",
"daniel",
"claude"
] |
p_0455461af0 | Make sure to run sudo when editing the crontab, otherwise you'll be modifying the user crontab instead of the system one. | [
"sudo",
"crontab"
] | [
"tech"
] | null | train | borrowed | danielrosehill/Tech-Sentences-For-ASR-Training | danielrosehill/Tech-Sentences-For-ASR-Training | MIT | https://huggingface.co/datasets/danielrosehill/Tech-Sentences-For-ASR-Training | [] | null | null | null | [
"daniel",
"daniel"
] |
p_047f0fc19b | Can you check why Radarr keeps grabbing the wrong release for that one movie? It's picking the low quality version again, and I'd rather it waited. | [
"Radarr"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_048e0e2ad6 | Our SLO requires ninety-nine point nine percent uptime, but we're burning through our error budget too quickly this quarter. | [
"SLO",
"error budget"
] | [
"tech"
] | null | train | borrowed | danielrosehill/Tech-Sentences-For-ASR-Training | danielrosehill/Tech-Sentences-For-ASR-Training | MIT | https://huggingface.co/datasets/danielrosehill/Tech-Sentences-For-ASR-Training | [] | null | null | null | [
"daniel",
"daniel"
] |
p_04d8180180 | Claude, set up LibreChat on the home server behind the reverse proxy, and connect it to both my OpenAI key and the local Ollama instance. | [
"LibreChat"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0503b39005 | Note to self, move the porch light logic out of the old YAML automations and into AppDaemon, so I can finally write it properly in Python. | [
"AppDaemon"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_052ff0f33d | I finally got the GGUF version running locally, and Kokoro sounds surprisingly good for the voice replies, even on the small server. | [
"GGUF",
"Kokoro"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"daniel",
"daniel"
] |
p_05305e3ca4 | The trace ID shows the request failed at the payment service, but the span data isn't appearing in Tempo. | [
"trace ID",
"span",
"Tempo"
] | [
"tech"
] | null | train | borrowed | danielrosehill/Tech-Sentences-For-ASR-Training | danielrosehill/Tech-Sentences-For-ASR-Training | MIT | https://huggingface.co/datasets/danielrosehill/Tech-Sentences-For-ASR-Training | [] | null | null | null | [
"daniel",
"daniel",
"daniel"
] |
p_055bbf85cf | For this distributed setup, we're considering DeepSpeed versus FSDP to handle model parallelism across our eight GPU cluster efficiently. | [
"DeepSpeed",
"FSDP",
"model parallelism"
] | [
"tech"
] | null | train | borrowed | danielrosehill/Tech-Sentences-For-ASR-Training | danielrosehill/Tech-Sentences-For-ASR-Training | MIT | https://huggingface.co/datasets/danielrosehill/Tech-Sentences-For-ASR-Training | [] | null | null | null | [
"daniel",
"daniel",
"daniel"
] |
p_056fa9cb1e | Please convert all the images in the uploads folder to WebP, keep the originals as a backup, and update the components so they use the new files. | [
"WebP"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_05d8c0c001 | English results on LibriSpeech should be viewed as an upper-bound estimate because LibriHeavy is a superset of LibriSpeech. | [
"LibriSpeech"
] | [
"tech"
] | null | train | mined-verbatim | hysts-bot-data/daily-papers | hysts-bot-data/daily-papers | CC0-1.0 (compilation; arXiv metadata, including abstracts, is CC0) | https://arxiv.org/abs/2604.01760 | [
"Chihiro Arata",
"Kiyoshi Kurihara"
] | 2026-04-03T00:00:00 | null | null | [
"daniel"
] |
p_05ebe8f459 | The JSON Web Token contains the user's role information, which our RBAC system uses to determine access permissions for each API endpoint. | [
"JSON Web Token",
"RBAC"
] | [
"tech"
] | null | train | borrowed | danielrosehill/Tech-Sentences-For-ASR-Training | danielrosehill/Tech-Sentences-For-ASR-Training | MIT | https://huggingface.co/datasets/danielrosehill/Tech-Sentences-For-ASR-Training | [] | null | null | null | [
"daniel",
"claude"
] |
p_05fa52aac6 | Hi Dana, I converted the onboarding guide to Markdown so it renders cleanly in the wiki. Take a look and tell me what's missing. | [
"Markdown"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"github-topics"
] |
p_0696626a8b | I rebuilt the calendar tool using FastMCP, and it took like twenty lines. Can you take a look before I hook it up to the agent? | [
"FastMCP"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"pypi"
] |
p_074d6031b7 | Can you check whether QUIC is getting blocked on the office network? The video calls keep falling back to something slower and I can't figure out why. | [
"QUIC"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_079d112ac5 | Just mounted the Reolink camera above the front door, but the night view is blurry. I'll check the focus and the settings tomorrow morning. | [
"Reolink"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_07ba3549d8 | We want to fine-tune a small model on our support emails, so could you send me a sample of the cleaned data? | [
"fine-tune"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"daniel"
] |
p_07c9783b86 | Please look at the Procfile in this repo and tell me why the Heroku deploy keeps failing, then fix it and push to a new branch. | [
"Heroku"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"github-topics"
] |
p_07fe48466f | Hi everyone, I was ecstatic after the recent Docker Unsloth release, so I packaged up a RunPod one-click template for everyone here. | [
"Docker",
"Unsloth",
"RunPod"
] | [
"tech"
] | null | train | mined-verbatim | pszemraj/LocalLLaMA-posts | pszemraj/LocalLLaMA-posts | ODC-BY-1.0 | https://www.reddit.com/r/LocalLLaMA/comments/1nyzzws/template_oneclick_unsloth_finetuning_on_runpod/ | [] | 2025-10-05T00:00:00 | null | null | [
"daniel",
"daniel",
"claude"
] |
p_08138be9fd | After the last update my laptop boots to a black screen, so I think SDDM is failing to start the login session. | [
"SDDM"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"daniel"
] |
p_084456c81f | Unlike approaches that adapt offline models through chunking or sliding windows, Voxtral Realtime is trained end-to-end for streaming, with explicit alignment between audio and text streams. | [
"Voxtral"
] | [
"tech"
] | null | train | mined-verbatim | hysts-bot-data/daily-papers | hysts-bot-data/daily-papers | CC0-1.0 (compilation; arXiv metadata, including abstracts, is CC0) | https://arxiv.org/abs/2602.11298 | [
"Alexander H. Liu",
"Andy Ehrenberg",
"Andy Lo",
"Chen-Yo Sun",
"Guillaume Lample",
"Jean-Malo Delignon",
"Khyathi Raghavi Chandu",
"Patrick von Platen",
"Pavankumar Reddy Muddireddy",
"Rohin Arora",
"Sanchit Gandhi",
"Sandeep Subramanian",
"Soham Ghosh",
"Srijan Mishra",
"Abhinav Rastog... | 2026-02-13T00:00:00 | null | null | [
"daniel"
] |
p_087bf67200 | Please go through the Forgejo repository, find the open branches that have not been touched in months, and list them so I can decide which ones to delete. | [
"Forgejo"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_08930bbff9 | The printer isn't showing up on the network, so check whether mDNS is enabled on the server and whether avahi is actually running. | [
"mDNS"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0895d20417 | I finally ordered the new workstation with a Ryzen processor, so please hold off on buying that other mini PC until I confirm the specs. | [
"Ryzen"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_094f10efe2 | Please add a devcontainer to this repository so anyone can open the project with the right Node version and database already set up. | [
"devcontainer"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0961ef634a | Please convert the docs page for the pricing section to MDX so we can embed the interactive calculator component right inside the text, and make sure the build still passes. | [
"MDX"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"github-topics"
] |
p_0962ae70e6 | Hi Dana, we're testing Jina AI for the reranking step, and so far the results look better than what we had before. Can we talk tomorrow? | [
"Jina AI"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_096f4827a6 | Hi Dana, I moved our internal docs into VitePress last night, so please look over the sidebar structure and tell me if anything seems out of order. | [
"VitePress"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_09c8e50a98 | Hi Dan, I tried Gemini CLI on the migration script yesterday and it caught a few bugs I missed. Happy to show you how it works. | [
"Gemini CLI"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_09c91938d6 | Check why training is using so much GPU memory on the server. If it isn't already enabled, turn on FlashAttention in the model config, then run a short test with the same batch size and compare memory use and speed. If the install fails because of the CUDA setup, don't try to fix everything. Just tell me exactly what's... | [
"FlashAttention"
] | [
"tech",
"long"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0a01252470 | Check the dnsmasq config on the Raspberry Pi and tell me why local hostnames stop resolving after the router hands out a new DHCP lease. | [
"dnsmasq"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0a2d1e2d22 | Current LLM-driven automated optimization methods narrowly focus on machine learning applications, such as PyTorch operator optimization, while overlooking broader domains like sparse matrix operations in scientific computing. | [
"LLM",
"PyTorch"
] | [
"tech"
] | null | train | mined-edited | hysts-bot-data/daily-papers | hysts-bot-data/daily-papers | CC0-1.0 (compilation; arXiv metadata, including abstracts, is CC0) | https://arxiv.org/abs/2603.07169 | [
"Yuxuan Han",
"Meng-Hao Guo",
"Zhengning Liu",
"Wenguang Chen",
"Shi-Min Hu"
] | 2026-03-10T00:00:00 | However, current LLM-driven automated optimization methods narrowly focus on machine learning applications, such as PyTorch operator optimization, while overlooking broader domains like sparse matrix operations in scientific computing. | fixed: stripped leading 'However,' | [
"daniel",
"daniel"
] |
p_0a715b0fd0 | Hi Dan, I'm sending the training logs now, so please open them in TensorBoard and let me know if the validation loss looks strange to you. | [
"TensorBoard"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0a755e2c5b | The deployment failed because ACR rejected our push, apparently the container registry has new authentication requirements we missed. | [
"ACR",
"container registry"
] | [
"tech"
] | null | train | borrowed | danielrosehill/Tech-Sentences-For-ASR-Training | danielrosehill/Tech-Sentences-For-ASR-Training | MIT | https://huggingface.co/datasets/danielrosehill/Tech-Sentences-For-ASR-Training | [] | null | null | null | [
"daniel",
"daniel"
] |
p_0a84f001c9 | Man, dealing with CommonJS modules in a modern Express app is such a pain when you want to use the latest JavaScript features. | [
"CommonJS",
"Express"
] | [
"tech"
] | null | train | borrowed | danielrosehill/Tech-Sentences-For-ASR-Training | danielrosehill/Tech-Sentences-For-ASR-Training | MIT | https://huggingface.co/datasets/danielrosehill/Tech-Sentences-For-ASR-Training | [] | null | null | null | [
"daniel",
"daniel"
] |
p_0a877dd34b | Claude, set up a ClickHouse table for the event logs, partition it by month, and make sure the ingest script batches its inserts instead of writing row by row. | [
"ClickHouse"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0a88e96c8f | What GPU is Moonshot Kimi K2 running on? | [
"GPU",
"Moonshot",
"Kimi"
] | [
"tech",
"numeric"
] | null | train | mined-verbatim | pszemraj/LocalLLaMA-posts | pszemraj/LocalLLaMA-posts | ODC-BY-1.0 | https://www.reddit.com/r/LocalLLaMA/comments/1m4t22z/what_gpu_is_moonshot_kimi_k2_running_on/ | [] | 2025-07-20T00:00:00 | null | null | [
"daniel",
"claude",
"daniel"
] |
p_0aa7e1dfb8 | Can you check whether the NPU on the new laptop actually gets used when we run the local transcription model, or does it fall back to the CPU? | [
"NPU"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0ab180fadd | Open btop on the server when you get a chance, because something is eating all the memory and I can't tell what. | [
"btop"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0ab44445c6 | Claude, please go through the repository for the living room sensors and check the ESPHome configuration for the new board. After that, look at why the CPU on the server spikes whenever the device reconnects. Don't change anything yet. First write me a short summary of what you found, including which services are runni... | [
"ESPHome",
"CPU"
] | [
"long",
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude",
"daniel"
] |
p_0ab5ce539f | Qwen keeps answering in Chinese when the prompt is short, so add an instruction for English. | [
"Qwen"
] | [
"tech"
] | null | train | generated | claude-opus-5.5 (Claude Code session) | null | null | null | [] | null | null | null | [
"daniel"
] |
p_0ac4746e8d | Remind me to check whether the library actually follows SemVer before we bump it, because last time a minor release broke half our build. | [
"SemVer"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"pypi"
] |
p_0affef4f2e | Please check the CrowdSec logs on the server and tell me which IP addresses got banned in the last day, and why. | [
"CrowdSec"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0b32ebc948 | Switched the test setup over to Vitest this morning and it's so much faster than Jest, since it just reuses the Vite config. | [
"Vitest"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0b74cdbbdc | Hi Dana, I finally flashed GrapheneOS onto my old Pixel, and so far it's running great. I'll tell you how the banking apps behave. | [
"GrapheneOS"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"github-topics"
] |
p_0ba474d563 | Analyzing CSV and structured data - RAG, MCP, tools, or plain old scripting? | [
"CSV",
"RAG",
"MCP"
] | [
"tech"
] | null | train | mined-verbatim | pszemraj/LocalLLaMA-posts | pszemraj/LocalLLaMA-posts | ODC-BY-1.0 | https://www.reddit.com/r/LocalLLaMA/comments/1m7mu6e/analyzing_csv_and_structured_data_rag_mcp_tools/ | [] | 2025-07-23T00:00:00 | null | null | [
"daniel",
"daniel",
"daniel"
] |
p_0bd85c590c | The rollout is phased across web and iOS, with 3 billion users eventually seeing Gemini touch their inboxes. | [
"iOS",
"Gemini"
] | [
"tech",
"numeric"
] | null | train | mined-edited | e1879/threads-english-tech-news-sft | e1879/threads-english-tech-news-sft | CC-BY-4.0 | https://www.threads.net/@ai.news.daily/post/DTS7BGxCJq9 | [] | 2026-01-09T00:00:00 | The rollout is phased across web + iOS, with 3B users eventually seeing Gemini touch their inboxes. | fixed: '+' and '3B' spelled for reading | [
"daniel",
"daniel"
] |
p_0bf816ab8c | Run the same prompt through DeepSeek and Qwen and compare the answers. | [
"DeepSeek",
"Qwen"
] | [
"tech"
] | null | train | generated | claude-opus-5.5 (Claude Code session) | null | null | null | [] | null | null | null | [
"daniel",
"daniel"
] |
p_0c7eae7b1c | I've been experimenting with multi-agent debate for the past few months — running structured council sessions across Claude, GPT, Gemini, DeepSeek, Grok, Kimi, and local models via Ollama. | [
"Claude",
"GPT",
"Gemini",
"DeepSeek",
"Grok",
"Kimi",
"Ollama"
] | [
"tech"
] | null | train | mined-verbatim | pszemraj/LocalLLaMA-posts | pszemraj/LocalLLaMA-posts | ODC-BY-1.0 | https://www.reddit.com/r/LocalLLaMA/comments/1qv4cv2/i_built_a_researchbacked_framework_for_running/ | [] | 2026-02-03T00:00:00 | null | null | [
"daniel",
"daniel",
"daniel",
"daniel",
"daniel",
"daniel",
"daniel"
] |
p_0c8b0893ff | I want you to build a small prototype using smolagents in this repository. Create one agent that can search our internal docs and answer questions about the smart home setup. Keep it very simple, with a single tool and a short system prompt. Use a local model through our existing server rather than a paid API. When it ... | [
"smolagents"
] | [
"tech",
"long"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0ca4735977 | Add a small script that turns on the Govee lights in the living room when the front door sensor opens after sunset. | [
"Govee"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0cb21aab3b | DeepSeek V3 is amazing, but I don't trust sending them my PII. | [
"DeepSeek",
"PII"
] | [
"tech",
"numeric"
] | null | train | mined-verbatim | pszemraj/LocalLLaMA-posts | pszemraj/LocalLLaMA-posts | ODC-BY-1.0 | https://www.reddit.com/r/LocalLLaMA/comments/1qqsto7/deepseek_v3_is_amazing_but_i_dont_trust_sending/ | [] | 2026-01-30T00:00:00 | null | null | [
"daniel",
"daniel"
] |
p_0cc5ba7914 | I want you to set up MkDocs in this repository so we have a proper documentation site. Create the config, add a home page, and move the existing readme sections into separate pages. Then add a navigation menu that follows the same order. Please use the Material theme, and make sure the site builds locally without any w... | [
"MkDocs"
] | [
"tech",
"long"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"github-topics"
] |
p_0cdd4dbbae | Claude, add a new client in Keycloak for the dashboard app, turn on the authorization code flow, and update the environment template with the realm and client settings. | [
"Keycloak"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"cncf"
] |
p_0cfe171bb2 | Please check my espanso config on this machine and add a new trigger that expands my work email signature. Then restart the service and make sure it loads without errors. | [
"espanso"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"daniel"
] |
p_0d13b941f0 | Can you rebase your branch on main before I review it? There are a bunch of conflicts and I'd rather not sort them out. | [
"rebase"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0d13f55674 | Can you check why the CalDAV sync keeps dropping my shared calendar on the phone? It only started happening after we moved the server. | [
"CalDAV"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0d2a3a4712 | Hi Dan, I moved the stylesheet build to PostCSS this morning, so please pull the latest branch before you touch any of the styling. | [
"PostCSS"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"github-topics"
] |
p_0d813b728b | We released Vision RL, allowing you to train Gemma 3, Qwen2.5-VL with GRPO free in our Colab notebooks. | [
"Gemma",
"GRPO",
"Colab"
] | [
"tech",
"numeric"
] | null | train | mined-verbatim | pszemraj/LocalLLaMA-posts | pszemraj/LocalLLaMA-posts | ODC-BY-1.0 | https://www.reddit.com/r/LocalLLaMA/comments/1nr4v7e/gptoss_reinforcement_learning_fastest_inference/ | [] | 2025-09-26T00:00:00 | null | null | [
"daniel",
"claude",
"claude"
] |
p_0dc827bf92 | Check whether the SDDM login screen still loads after the update, and then rerun the benchmark so we can compare the WER. | [
"SDDM",
"WER"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"daniel",
"daniel"
] |
p_0dd09a64dc | Hey, did you manage to run OLMo on the new GPU box yet? I'm curious how the memory usage looks compared to the last model. | [
"OLMo"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0e1210ba7c | Hi Dana, I've shared the updated Figma file with you. Please leave your comments on the onboarding screens before Thursday so we can lock the design. | [
"Figma"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"github-topics"
] |
p_0e1542fb21 | We're planning to replace our monolithic application with microservices deployed on Cloud Run, using Pub/Sub for asynchronous communication between services. | [
"Cloud Run",
"Pub/Sub"
] | [
"tech"
] | null | train | borrowed | danielrosehill/Tech-Sentences-For-ASR-Training | danielrosehill/Tech-Sentences-For-ASR-Training | MIT | https://huggingface.co/datasets/danielrosehill/Tech-Sentences-For-ASR-Training | [] | null | null | null | [
"daniel",
"daniel"
] |
p_0e27367f0f | Hi Dan, I ran htop on the new server this morning and noticed one Node process eating almost all the memory, so please take a look. | [
"htop"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0e296a07c1 | Across GPT-2 Small, Gemma 2 2B, and Qwen 3.5 2B Base, ICALens efficiently recovers compact, human-interpretable directions without per-layer gradient-based dictionary training. | [
"GPT",
"Gemma",
"Qwen"
] | [
"tech",
"numeric"
] | null | train | mined-verbatim | hysts-bot-data/daily-papers | hysts-bot-data/daily-papers | CC0-1.0 (compilation; arXiv metadata, including abstracts, is CC0) | https://arxiv.org/abs/2606.11722 | [
"Sida Liu",
"Feijiang Han"
] | 2026-06-11T00:00:00 | null | null | [
"daniel",
"daniel",
"daniel"
] |
p_0e3de226c6 | Please look through the repository and add support for a LoRA adapter loading step, then make sure the webhooks still fire correctly after training finishes. | [
"LoRA",
"webhooks"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"daniel",
"daniel"
] |
p_0e641c0ecd | We evaluate InternAgent-1.5 on scientific reasoning benchmarks such as GAIA, HLE, GPQA, and FrontierScience, and the system achieves leading performance that demonstrates strong foundational capabilities. | [
"HLE",
"GPQA"
] | [
"tech",
"numeric"
] | null | train | mined-verbatim | hysts-bot-data/daily-papers | hysts-bot-data/daily-papers | CC0-1.0 (compilation; arXiv metadata, including abstracts, is CC0) | https://arxiv.org/abs/2602.08990 | [
"Shiyang Feng",
"Runmin Ma",
"Xiangchao Yan",
"Yue Fan",
"Yusong Hu",
"Songtao Huang",
"Shuaiyu Zhang",
"Zongsheng Cao",
"Tianshuo Peng",
"Jiakang Yuan",
"Zijie Guo",
"Zhijie Zhong",
"Shangheng Du",
"Weida Wang",
"Jinxin Shi",
"Yuhao Zhou",
"Xiaohan He",
"Zhiyin Yu",
"Fangchen Yu... | 2026-02-10T00:00:00 | null | null | [
"claude",
"claude"
] |
p_0e64ce5152 | Claude, please set up openWakeWord on the living room Raspberry Pi so it listens for our custom phrase, and log every detection with a timestamp. | [
"openWakeWord"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0e8d218834 | Bought three Sonoff switches for the balcony lights, but I still need to check whether they pair with the Zigbee hub before I drill anything. | [
"Sonoff"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0e94b9c276 | Just saw that Cerebras is serving that open model at crazy speeds. I want to test their API for the voice assistant tonight. | [
"Cerebras"
] | [
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0e9939ecfd | Look at the SnapRAID configuration on the home server and tell me when the last sync actually finished. I want to know if any of the data drives are throwing errors. Don't run a fix yet. First show me the parity file locations and the content files, and check that they are not all on the same disk. After that, write me... | [
"SnapRAID"
] | [
"tech",
"long"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"claude"
] |
p_0eaad2ad9e | Okay, so I finally got the model running on the home server. I used heavier quantization to fit it in memory, and it is still pretty accurate. Next I need to add a crontab entry to restart the service every night, because it leaks memory after a day or two. Remind me to check the logs tomorrow morning and see whether t... | [
"quantization",
"crontab"
] | [
"long",
"tech"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"daniel",
"daniel"
] |
p_0ec8e3d715 | I want to try rewriting the small log parser in Zig. Start by creating a new folder next to the existing Python version, and set up a minimal build there. Don't touch the current code yet. Port just the part that reads lines and splits the timestamp from the message. Then write a simple benchmark so we can compare spee... | [
"Zig"
] | [
"tech",
"long"
] | null | train | generated | anthropic/claude-sonnet-5.5 | null | null | null | [] | null | null | null | [
"github-topics"
] |
End of preview. Expand in Data Studio
AI and Tech Vocabulary Sentence Bank
English sentences dense in AI and developer vocabulary: model and lab names (Claude, Qwen, DeepSeek, Kimi, GLM), benchmarks (SWE-bench, GPQA, ARC-AGI), inference tooling (vLLM, llama.cpp, Ollama, Unsloth), cloud and DevOps tools. Newer names are favoured.
Website: hebrish.app: get the Hebrish keyboard for Android and the desktop apps, see the results, and find every model, dataset and demo in one place.
These are sentences to read aloud for fine-tuning a speech-recognition model, not recordings. They feed a targeted fine-tune of NVIDIA Parakeet TDT 0.6B v2 (current model). A capture app shows these sentences one at a time to be read aloud; every take lands in Read-Aloud-Voice-Sentences and is copied into the matching audio subset. The recordings of these sentences are in Tech-Sentences.
This dataset gathers sentences from several places into one file for my own training, and it is reproduced here for convenience. Every row says where it came from, who wrote it and how it was changed. Credit for upstream sentences belongs to their authors (see "Attribution").
Size
1806 train, 227 test (2033 total). The test split is term-disjoint:
no target term in test appears in train, so an evaluation on it measures generalisation,
not memorisation.
Derivation: generated 1356, mined-verbatim 341, borrowed 225, mined-edited 111.
Sentence authors: anthropic/claude-sonnet-5.5 1335, pszemraj/LocalLLaMA-posts 232, danielrosehill/Tech-Sentences-For-ASR-Training 225, hysts-bot-data/daily-papers 161, e1879/threads-english-tech-news-sft 59, claude-opus-5.5 (Claude Code session) 21.
Columns
| Column | Meaning |
|---|---|
id |
sentence id: a hash of the text (same id as the capture app and the audio datasets' prompt_id) |
text |
the sentence, exactly as read. Punctuated and cased; numbers in digits for versions, sizes, percentages and money |
target_terms |
the words the sentence exists to teach |
term_origins |
for each target term, who proposed it: daniel (from my own repos and notes, or I asked for it), claude (an AI agent proposed it), or the public list it came from (openrouter, github-topics, cncf, pypi, stackoverflow) |
categories |
tech, linux, smart_home, homophone, numeric, … |
homophone_set |
homophone rows only: the set the context disambiguates |
split |
train or test |
derivation |
mined-verbatim (taken unchanged from a public dataset), mined-edited (taken, then lightly fixed for casing, digits or readability), generated (written by an LLM for this project, then reviewed by a second model), borrowed (taken whole from an earlier sentence set) |
sentence_author |
the upstream dataset, or the model that wrote the sentence |
source_dataset, source_licence, source_url, source_authors, source_date |
upstream provenance for mined and borrowed rows. source_url links the paper or post; source_authors lists arXiv authors (Reddit usernames are deliberately not copied) |
original_text |
the upstream text before editing (mined-edited rows) |
review_note |
the reviewer's note when a sentence was fixed |
Attribution and licence
Licensing is per row (source_licence). Rows without one were written for this project and
are CC-BY-4.0.
| Upstream | Licence |
|---|---|
| danielrosehill/Tech-Sentences-For-ASR-Training | MIT |
| e1879/threads-english-tech-news-sft | CC-BY-4.0 |
| hysts-bot-data/daily-papers | CC0-1.0 (compilation; arXiv metadata, including abstracts, is CC0) |
| pszemraj/LocalLLaMA-posts | ODC-BY-1.0 |
Mined sentences were selected by scripts/mine_tech_sentences.py in the project repository:
they are 8–30 words long, contain at least one named tech term, and come from newer posts first.
An AI reviewer then kept or dropped each one and fixed casing and numbers.
Updating
Upstream sources can be re-mined, and new sentences are added as the project grows. Ids are content hashes, so an edited sentence gets a new id.
- Downloads last month
- 30