Datasets:
id string | file_name string | date timestamp[s] | time string | title string | category string | tags list | speaker string | language string | transcript string | for_claude bool | claude_done_at timestamp[s] | claude_reviewed_at timestamp[s] | claude_decision string | claude_category string | claude_actioned bool | claude_log string | transcript_revised string | transcript_revised_at timestamp[s] | repo_copy string | duration_s float64 | sample_rate int64 | bit_depth int64 | audio_format string | source string | transcript_model string | recorded_at string |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2026-10-26-152336 | audio/2026-10-26_improvement-note-1_15-23-36.mp3 | 2026-10-26T00:00:00 | 15:23:36 | The Ideas dataset (meta-idea) | Dataset design | [
"ideas",
"meta-idea",
"dataset-design",
"capture-app"
] | Daniel Rosehill | en | This is a voice note I'm recording while I'm working on my voice note collection app for training data for my fine tunes, which is essentially whenever I have a spare moment, I'm just sending in some clips here and there. I've been trying to mix up recordings that were done at home, in different rooms, friends' places,... | false | null | null | null | null | false | null | null | null | null | null | null | null | null | null | null | null |
2026-10-26-153216 | audio/2026-10-26_improvement-note-2_15-32-16.mp3 | 2026-10-26T00:00:00 | 15:32:16 | Annotation design and the training plan | Annotation and training | [
"annotation",
"noise",
"background-conversation",
"background-music",
"training-plan",
"vocab-queue"
] | Daniel Rosehill | en | All right, so this is the second part of my notes for the dataset collection improvement that I was in the middle of there. If I go in some settings where I'm out and about and I'm collecting my voice notes, which are going to be tagged according to whether they're tech vocab, code-switching, as the case may be, probab... | false | null | null | null | null | false | null | null | null | null | null | null | null | null | null | null | null |
2026-10-26-155219 | audio/2026-10-26_improvement-note-3_15-52-19.mp3 | 2026-10-26T00:00:00 | 15:52:19 | Technical ideas to explore | Technical ideas | [
"noisy-speech",
"cross-lingual",
"speaker-verification",
"feasibility"
] | Daniel Rosehill | en | I have an idea maybe we can categorize these notes. I'm speaking to you, Claude Code, or I will be speaking to you, this is for your attention. We can categorize this dataset of ideas and decompose them into categories, and maybe each one can be decomposed into sections, but an idea came to me in the grocery store ther... | false | null | null | null | null | false | null | null | null | null | null | null | null | null | null | null | null |
20261030T104948Z_fhz3oh | audio/20261030T104948Z_fhz3oh.flac | 2026-10-30T00:00:00 | 10:49:48 | Room tone | Dataset design | [] | Daniel Rosehill | en | So, one quick thing to improve in my little capture app here is the room tone capture because it's not just room tone is useful, but I think background tone is more important. So, I'm going to capture this manually offline. 60 seconds of unfiltered, unprocessed of Jaffa Street as it is as I'm recording these this morni... | false | null | null | null | null | false | null | null | null | null | 230.373 | 48,000 | 24 | flac | capture-app | google/gemini-3.5-flash | 2026-10-30T10:49:48.196Z |
20261031T115507Z_e5j62q | audio/20261031T115507Z_e5j62q.flac | 2026-10-31T00:00:00 | 11:55:07 | Good samples | Other | [] | Daniel Rosehill | en | All right, this is just a note as I'm recording some samples. So, one that just came up for me was 900 Mbps, which of course one would naturally say 900 megabytes per second. Or in maybe more realistic internet speeds these days, 1.5 Gbps. I mean, some people would say Gbps, but I think more commonly people would say g... | true | 2026-10-31T00:00:00 | 2026-10-31T00:00:00 | act | task-prompts | true | 2026-10-31: Reviewed: numerics already has 20 units prompts (incl. 900 Mbps); missing ms/latency, Gbps, MB/s, ounces, ml, tbsp, km/h. Plan ~60 prompts in data/task-prompts/numerics.json. Waiting on Daniel: abbreviated vs word targets (open question 37) and Mbps = megabits.
2026-10-31: Daniel settled: tech units abbrevi... | null | null | null | 97.125 | 48,000 | 24 | flac | capture-app | google/gemini-3.5-flash | 2026-10-31T11:55:07.542Z |
20261031T121138Z_0mj78n | audio/20261031T121138Z_0mj78n.flac | 2026-10-31T00:00:00 | 12:11:38 | Stop showing type | Other | [
"capture-app"
] | Daniel Rosehill | en | All right, so this is a note for the capture app. So when I'm doing the sentences and recording them, so right now there's a lot of Israeli trade names that I'm not familiar with and for these I'm going to have to like look them up or maybe get someone else to do it, like a doctor, if I can coerce someone or find someo... | true | 2026-10-31T00:00:00 | 2026-10-31T00:00:00 | act | capture-app | true | 2026-10-31: Reviewed: wants a 'Stop showing <category>' choice under Skip (e.g. Israeli medical brand names), hidden until unhidden, with a Hidden list at the bottom of Tasks to restore. Not built yet; proposed to Daniel.
2026-10-31: Built in capture app release 109: Skip now offers 'Stop showing <type>' (bank categori... | null | null | null | 68.837 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-10-31T12:11:38.844Z |
20261031T122154Z_0vpjqq | audio/20261031T122154Z_0vpjqq.flac | 2026-10-31T00:00:00 | 12:21:54 | Word recurrence | Other | [] | Daniel Rosehill | en | Okay, so I'm recording at the moment and I feel like there are certain words that come up, so I've been doing samples with KDE Plasma in them today, and, you know, it's not a hugely important piece of vocab, I mean it's useful, but I question exactly how many samples we need and whether it's worth, especially as a larg... | true | 2026-10-31T00:00:00 | 2026-10-31T00:00:00 | act | dataset | true | 2026-10-31: Built: ASR-Training data/term-importance.json rates all 5,515 bank target terms 1-5 (first pass by Claude subagents; anchors SSH key 5, KDE Plasma 3), mapped to 3/5/10/15/20 takes, unrated 3. Capture app release 109 serves a sentence last once every target term in it has reached its target; lexicon_saturati... | null | null | null | 107.237 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-10-31T12:21:54.233Z |
20261031T122710Z_8stpot | audio/20261031T122710Z_8stpot.flac | 2026-10-31T00:00:00 | 12:27:10 | Untitled note | Capture app | [] | Daniel Rosehill | en | Okay, so, one capture app fix here to do is in the idea notes, so we have subtabs for idea, background noise, and room tone. Let's just have like idea, background noise, room. I'm just trying to keep it like smaller, those tabs. And I don't need these descriptions: a voice note about the project, what to build next, be... | true | 2026-10-31T00:00:00 | 2026-10-31T00:00:00 | act | capture-app | true | 2026-10-31: One media control bar shared by Idea, Background noise and Room (Pause, Stop, Retake | Record | Save, Delete); Recorder release 110, sprint planning/sprints/2026-10-31-ideas-review.md F3 | null | null | null | 60.661 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-10-31T12:27:10.701Z |
20261031T122817Z_mwo1c0 | audio/20261031T122817Z_mwo1c0.flac | 2026-10-31T00:00:00 | 12:28:17 | Improving | Capture app | [] | Daniel Rosehill | en | One other thing for the capture app, for the idea notes recording things, so it says after I, you know, stop it, it says that it's uploading. And maybe just have a little success thing or uploaded. And I feel like it should be polling more frequently because it was hanging there for like 20 seconds, which I feel probab... | true | 2026-10-31T00:00:00 | 2026-10-31T00:00:00 | act | capture-app | true | 2026-10-31: Upload status listed under the bar; uploading hang fixed (pump released its lock before refreshing); category icons; shorter tab labels; Recorder release 110, sprint ideas-review F3 | null | null | null | 43.775 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-10-31T12:28:17.306Z |
20261031T122919Z_b6mny9 | audio/20261031T122919Z_b6mny9.flac | 2026-10-31T00:00:00 | 12:29:19 | Recording | Capture app | [] | Daniel Rosehill | en | And for this is also feedback on the um ideas tab in the recording app. So background noise room tone. Let's have a consistent media element throughout. Again same problem as before, and we just have one button. Let's standardize it to make it aligned with the sample recording in the um idea sorry in the main recording... | true | 2026-10-31T00:00:00 | 2026-10-31T00:00:00 | act | capture-app | true | 2026-10-31: Slow transcription traced to default reasoning on Gemini Flash, sequential 60 s chunks and two commits per note; now one request with minimal reasoning, a 16 kbps Opus copy instead of WAV, one commit: 5-7 s for a 2 minute note. Stray last sentences came from a tiny tail chunk, removed. Recorder release 110,... | null | null | null | 27.119 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-10-31T12:29:19.555Z |
20261031T123221Z_sux6i5 | audio/20261031T123221Z_sux6i5.flac | 2026-10-31T00:00:00 | 12:32:21 | Dataset | Capture app | [] | Daniel Rosehill | en | So in the idea section of the app, I presume that my recordings are being uploaded immediately? We're not waiting on the general queue. That should be the case if not, and because these ones are, it's kind of a quick sync, and the other ones are in the background. Regarding the general push to using the Parquet format,... | true | 2026-10-31T00:00:00 | 2026-10-31T00:00:00 | act | capture-app | true | 2026-10-31: Answer: Ideas notes upload immediately, independent of the take queue. Push by data size and slower child-dataset updates recorded as ideas, not actioned: planning/ideas/IDEA-0002.md, IDEA-0003.md | null | null | null | 120.234 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-10-31T12:32:21.865Z |
20261031T123640Z_d7zxpk | audio/20261031T123640Z_d7zxpk.flac | 2026-10-31T00:00:00 | 12:36:40 | Untitled note | Dataset design | [] | Daniel Rosehill | en | Okay, so today I was doing the readings and I came across Abarbanel Street, it's a bit of a tongue twister, Abarbanel Street in Jerusalem, which is a real street, and I was thinking that's actually a good one to have because definitely a use case here would be Waze, like navigation, where that's often actually quite sp... | true | 2026-10-31T00:00:00 | 2026-10-31T00:00:00 | act | task-prompts | true | 2026-10-31: Israeli streets round 2: 42 streets named after people, from the national street list; 212 phone prompts after a review rejected 6. The most frequent names (Jabotinsky, Herzl and others) already had sentences and were left out. ASR-Training 21bd9ab, Recorder release 111, sprint ideas-review F6 | null | null | null | 179.333 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-10-31T12:36:40.271Z |
20261031T123810Z_g01hqv | audio/20261031T123810Z_g01hqv.flac | 2026-10-31T00:00:00 | 12:38:10 | Project organization | Other | [] | Daniel Rosehill | en | So, I'm sending in just a little batch of notes today, using the ideas tab which is coming along. And what I'd like to do, this is a message for the attention of Dear Claude. In fact, maybe that filter is bad. We can take it off because probably most things are going to be for Claude. At least it should be default turn... | true | 2026-10-31T00:00:00 | 2026-10-31T00:00:00 | act | capture-app | true | 2026-10-31: For Claude ticked by default and after each save (Recorder release 111, F5). Skill claude-ideas now files every reviewed note into the Recorder planning tree: sprints for work done, features/to-do for backlog, ideas for not-yet (ASR-Training 21bd9ab, D1) | null | null | null | 77.274 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-10-31T12:38:10.893Z |
20261031T125226Z_2yjpzn | audio/20261031T125226Z_2yjpzn.flac | 2026-10-31T00:00:00 | 12:52:26 | Audio and | Other | [] | Daniel Rosehill | en | Here is another audio or a vocab set to review, which is basically common file names. So, this is one that I'll often see speech to text struggle with, especially when using Claude Code or it'll reach for very different versions every time. So on Kdenlive video editor, I think it's VAAPI, ROCm support, AMD. But the mor... | true | 2026-10-31T00:00:00 | 2026-10-31T00:00:00 | act | vocabulary | true | 2026-10-31: Tech pack media-formats: 35 terms (WebP, SVG, MP4, MKV, M4A, HEIC, VAAPI, ROCm, Kdenlive, transcode, downscale, 1080p, 16:9, 4:3, codecs), 154 sentences framed as instructions to an agent. Backticks for commands recorded as open question 51, not actioned. ASR-Training 21bd9ab, Recorder release 111, sprint i... | null | null | null | 230.661 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-10-31T12:52:26.652Z |
20261101T072744Z_v86cqs | audio/20261101T072744Z_v86cqs.flac | 2026-11-01T00:00:00 | 07:27:44 | Mobile recording gear and noise testing | Other | [] | Daniel Rosehill | en | Okay, so this is just a recording for not actionable necessarily for vlogs, maybe just for documentation. What I want to record is I would call of my mobile recording gear, kit out at the moment, because I'm trying to the model is getting really good, which is which is brilliant, because every day, I get to use it and ... | true | null | 2026-11-01T00:00:00 | act | capture-app | false | 2026-11-01: Headset model, link and Bluetooth codec per take: Recorder planning FEAT-0004 (release 117 already saves the input identity and lists Yealink BH71). Noise conditions across all tasks noted there. | null | null | null | 250.234 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T07:27:44.415Z |
20261101T074009Z_jein0w | audio/20261101T074009Z_jein0w.flac | 2026-11-01T00:00:00 | 07:40:09 | Walking mode for dictation app | Capture app | [] | Daniel Rosehill | en | Okay, so this is a note for an idea for the dictation app that maybe I recorded the last day. I can't remember if I mentioned this big longer voice note dump, but in case I didn't, here's it in isolation. So, I want to record more samples when I'm out and about, meaning walking. It's nice to take walks, and I also dict... | true | null | 2026-11-01T00:00:00 | act | capture-app | false | 2026-11-01: Filed as Recorder FEAT-0005 (walking tag), FEAT-0006 (practice mode for the gamepad), FEAT-0007 (walking mode: TTS, beep, record; gamepad first, voice commands later). | null | null | null | 331.471 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T07:40:09.468Z |
20261101T074552Z_hi6alv | audio/20261101T074552Z_hi6alv.flac | 2026-11-01T00:00:00 | 07:45:52 | Microphone tagging and headset audio testing | Capture app | [] | Daniel Rosehill | en | Okay, so, just for information, in case the app doesn't work for the microphone tagging, I am recording this on the microphone. The last sentence I read was "The square room is about 14 m²." That was the last transcriptive thing. And I'm just going to do a quick touch test, tapping it three times to validate the microp... | true | null | 2026-11-01T00:00:00 | act | capture-app | false | 2026-11-01: Mic-in-use indicator: FEAT-0003. Headset list (Poly 5200, Soundcore, 2.4G dongle): FEAT-0004, waiting on the device names. | null | null | null | 119.295 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T07:45:52.449Z |
20261101T074936Z_03knx4 | audio/20261101T074936Z_03knx4.flac | 2026-11-01T00:00:00 | 07:49:36 | Bluetooth microphone validation and preferences | Capture app | [] | Daniel Rosehill | en | All right. Other ideas for the voice recognition. So here is the scenario. Let's say that I'm out and I'm right now I'm using a USB a Bluetooth microphone and I want to make sure am I Is this actually getting into the recording? From previous experience of using USB using Bluetooth mics, on Android, not recognized, or ... | true | null | 2026-11-01T00:00:00 | act | capture-app | false | 2026-11-01: Mic check (live level, 10 s test clip) and Bluetooth-preferred input with an internal-mic option: Recorder FEAT-0003. | null | null | null | 150.138 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T07:49:36.703Z |
20261101T075200Z_c5k3nm | audio/20261101T075200Z_c5k3nm.flac | 2026-11-01T00:00:00 | 07:52:00 | Recording unprocessed audio for training datasets | Capture app | [] | Daniel Rosehill | en | One other idea for the voice app is that I was using the other day ASR mic a great, ASR MIC, excellent Android app voice recorder, I think the best one really, and it had an unprocessed mode. So we were initially recording the recording these voice clips down sampled for whatever Parakeet natively uses. And then I real... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | covered | capture-app | false | 2026-11-01: Takes are already 48 kHz / 24-bit FLAC with echo cancellation, noise suppression and AGC off; downsampling stays in training. Native-app UNPROCESSED source kept as Recorder IDEA-0004. | null | null | null | 136.426 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T07:52:00.755Z |
20261101T075340Z_7ij4bo | audio/20261101T075340Z_7ij4bo.flac | 2026-11-01T00:00:00 | 07:53:40 | Capture app batch upload and controls | Capture app | [] | Daniel Rosehill | en | One more note for Claude regarding the capture app tiny things. So, I'm recording a batch of ideas this morning, a batch of voice notes, which is probably how I'll actually use this feature the most. And I don't like the auto advance. I don't think it's actually very, it's not really appropriate in this particular cont... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | capture-app | true | 2026-11-01: Ideas: no auto-advance, confirm before upload, upload in batches of 5 with a manual Upload queue: Recorder FEAT-0001.
2026-11-01: Release 121: Save stops into review and Save again keeps the note, with no auto-advance. Notes upload in batches of 5 (or 1 or 10) and as one dataset commit per batch, and Upload... | null | null | null | 96.97 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T07:53:40.801Z |
20261101T075418Z_qa2ana | audio/20261101T075418Z_qa2ana.flac | 2026-11-01T00:00:00 | 07:54:18 | Disabling auto-advance after voice notes | Capture app | [] | Daniel Rosehill | en | So, as the last voice note did the auto-advancing, to be clear, what I would like is not have that functionality. When I click save next at the voice note stage, it can progress to the confirmation dialogue and it's up to you to confirm, so it's just like one extra button press step. And then if I do stop, then I can m... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | capture-app | true | 2026-11-01: Same as 7ij4bo: Recorder FEAT-0001.
2026-11-01: Release 121: no auto-advance after Save on the Ideas page. | null | null | null | 35.663 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T07:54:18.791Z |
20261101T075720Z_81h9yq | audio/20261101T075720Z_81h9yq.flac | 2026-11-01T00:00:00 | 07:57:20 | Handling niche Hebrew pronunciation variants in vocabulary | Capture app | [] | Daniel Rosehill | en | All right. So, one note, this one's actually regarding vocab. So, we should have a category in the ideas thing for for vocab because a lot of my comments They're probably be either data set design, which to be honest, vocab vocab is more a separate a separate concern. And I just want to add as a note that I saw a readi... | true | null | 2026-11-01T00:00:00 | act | dataset | false | 2026-11-01: Regional readings (Aron Koidesh) as a pronunciation-variant category: Recorder FEAT-0009. Vocabulary category for Ideas: FEAT-0002. Recurrence of trained Hebrew words is covered by the per-term take targets (release 109). | null | null | null | 109.978 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T07:57:20.383Z |
20261101T075834Z_ap6pyp | audio/20261101T075834Z_ap6pyp.flac | 2026-11-01T00:00:00 | 07:58:34 | Bluetooth mic capture app waveform bug | Capture app | [] | Daniel Rosehill | en | Let's add another category for bugs in the capture app, like maybe we have a subcategory under capture app ideas, bugs, and we can triage accordingly. So here's one: I was just recording a sentence, I went back to playback, and the waveform did not display the audio captured. I think it is a Bluetooth-related error bec... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | capture-app | true | 2026-11-01: Missing waveform after a Bluetooth take: Recorder BUG-0001. Capture app bug category: FEAT-0002.
2026-11-01: Release 121: the Bluetooth takes' audio was fine (all peaks -8.9 to -1 dBFS). The waveform now redraws when the canvas had no width or the page returns, and play-last shows the waveform of an auto-sa... | null | null | null | 48.815 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T07:58:34.972Z |
20261101T080047Z_izv51z | audio/20261101T080047Z_izv51z.flac | 2026-11-01T00:00:00 | 08:00:47 | Background noise from a power washer | Capture app | [] | Daniel Rosehill | en | I just thought I'd record a voice a a quick note uh talking about um background noise presence. So, just just when I was about to uh wax lyrical about how relaxing it is to sit on a you know, a bench and just record some of these sentences, some guy rocks up with uh one of those delightful power washers. Uh so that was... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | drop | other | false | 2026-11-01: Information only (power washer in the background); nothing to do. | null | null | null | 58.186 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T08:00:47.208Z |
20261101T080231Z_iht5pg | audio/20261101T080231Z_iht5pg.flac | 2026-11-01T00:00:00 | 08:02:31 | Walking icon for recording annotations | Capture app | [] | Daniel Rosehill | en | Oh okay, one more annotation that I'd like to have. And what I'm thinking about for this app is that to be able to use these effectively we are I'm on my Android, small screen. So icons, icons reign supreme. So what would be a helpful one would be So, now I'm walking. And it would be good to have a little I'm envisioni... | true | null | 2026-11-01T00:00:00 | act | capture-app | false | 2026-11-01: Walking icon toggle with an activity column: Recorder FEAT-0005, with a backfill note for the walk after this note. | null | null | null | 79.711 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T08:02:31.184Z |
20261101T080548Z_7g57nf | audio/20261101T080548Z_7g57nf.flac | 2026-11-01T00:00:00 | 08:05:48 | Walking noise sample category ideas | Capture app | [] | Daniel Rosehill | en | Maybe in the ideas thing, I'd like to have a sample for a option category for just like noise sample, or conflict sample. So, what I want to do is record a quick sample when I'm walking as I am now and using internal microphone. I can do that in the noise floor settings, but it would be good to be able to do that, so I... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | capture-app | true | 2026-11-01: Noise sample category for Ideas: Recorder FEAT-0002.
2026-11-01: Release 121: Noise sample category on the Ideas page. | null | null | null | 34.319 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T08:05:48.636Z |
20261101T081620Z_ch4w11 | audio/20261101T081620Z_ch4w11.flac | 2026-11-01T00:00:00 | 08:16:20 | Headphone testing for dictation and comfort | Capture app | [] | Daniel Rosehill | en | All right, going to just make a note today as I'm going through all my headphones. I tried to do a few takes with a bunch of them, and uh specifically eager to try the Poly um 5200 to see how it fares against the this Yealink microphone, as these are my two quality over-the-ear dictation microphones. And also, for comf... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | drop | other | false | 2026-11-01: Marked by Daniel as not for Claude (headphone comfort). | null | null | null | 38.063 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T08:16:20.530Z |
20261101T081736Z_w0iaqq | audio/20261101T081736Z_w0iaqq.flac | 2026-11-01T00:00:00 | 08:17:36 | Option to skip trailing periods | Capture app | [] | Daniel Rosehill | en | One frustration that I find in dictation, and maybe this is something This isn't for the capture app so much as for Hebrew, so we should probably have a section for app ideas distinct the capture app versus the actual app is, commonly, let's say that I'm typing into my inventory system, for example, and I'm quickly usi... | true | null | 2026-11-01T00:00:00 | act | other | false | 2026-11-01: Skip trailing period at insertion: Hebrish-App-Planning backlog L3. Hebrish app category for Ideas: Recorder FEAT-0002. | null | null | null | 41.015 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T08:17:36.207Z |
20261101T081852Z_jsbjw2 | audio/20261101T081852Z_jsbjw2.flac | 2026-11-01T00:00:00 | 08:18:52 | Automating note processing with agents | Capture app | [] | Daniel Rosehill | en | A stretch idea, and this would be very slick, is using the upload, the ideas upload queue triggers a webhook. Claude or an agent running in an agent workspace with the repo gets kicked off with a with a skill like Process Notes and it runs through that, pushes updates. So I literally just send them and the agent gets i... | true | null | 2026-11-01T00:00:00 | act | other | false | 2026-11-01: Ideas upload triggers an agent running claude-ideas: Recorder IDEA-0005, after FEAT-0001 batching. | null | null | null | 30.303 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T08:18:52.839Z |
20261101T082032Z_34n3rj | audio/20261101T082032Z_34n3rj.flac | 2026-11-01T00:00:00 | 08:20:32 | Expanding speech recognition vocabulary using Google Maps | Capture app | [] | Daniel Rosehill | en | I'd also like to have an NTS Note to Self category for ideas because some some just come to me on the fly. So there's a few to add to the vocab list. Jerusalem businesses, Maoz Falafel store, and Piskat HaFalafel. I think I was just recording those. The easiest way would probably be especially for street names as well ... | true | null | 2026-11-01T00:00:00 | ask-daniel | vocabulary | false | 2026-11-01: Maoz Falafel added (pack jewish-honorifics, Recorder release 119). Piskat HaFalafel: no source found under that name, spelling needed. Map sweep of his places: Recorder FEAT-0008. Note to self category: FEAT-0002. | null | null | null | 40.453 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T08:20:32.229Z |
20261101T082257Z_eyueby | audio/20261101T082257Z_eyueby.flac | 2026-11-01T00:00:00 | 08:22:57 | Adding Jewish honorifics to speech recognition vocabulary | Capture app | [] | Daniel Rosehill | en | One thing to add vocabulary for is what you could call Jewish honorifics. So, we have we had olav ha-sholom and we should also have aleha hashalom and harav hagaon, harav, shlit"a, Hashem yikom damam, all the various things that might be encountered. Yimakh shemo, yimakh shemam, these little things people drop in. Stil... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | vocabulary | true | 2026-11-01: Pack jewish-honorifics: 12 honorifics (aleha hashalom, zichronah livracha, harav hagaon, shlita, zatzal, zecher tzadik livracha, Hashem yikom damam/damo, yimach shemo/shemam, lehavdil, ad me'ah ve'esrim), 39 sentences, Recorder release 119. Shiurim as a transcribed-speech source not yet in the training-set ... | null | null | null | 86.789 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T08:22:57.699Z |
20261101T085636Z_bkudlg | audio/20261101T085636Z_bkudlg.flac | 2026-11-01T00:00:00 | 08:56:36 | Audio dataset backup tool | Capture app | [] | Daniel Rosehill | en | All right. So, very nice. I came back home and saw that Claude has been very busy working on the updates to the app and all the other stuff that I mentioned today. So, one other thing, one thought that I have is regarding the data format for storing the audio data. So, one idea, this isn't for immediate implementation,... | true | 2026-11-03T00:00:00 | 2026-11-03T00:00:00 | act | other | true | 2026-11-03: Built 2026-11-01: public repo Audio-Dataset-Backup (discover HF audio datasets, snapshot, pack into disc-sized volumes, verify, restore). planning/sprints/2026-11-03-snappy-advance-and-word-presets.md | null | null | null | 166.981 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T08:56:36.294Z |
20261101T090352Z_1197gc | audio/20261101T090352Z_1197gc.flac | 2026-11-01T00:00:00 | 09:03:52 | Telegram bot for vocabulary collection | Capture app | [] | Daniel Rosehill | en | All right, here is one tool I'd like to add today, a Telegram bot so Hannah and I can use this. And what I'd like it to do is, I asked Hannah, who is also a big dictation fan, and she's an architect, my wife. Are there any words like, you know, Revit, all the vocab she uses every day? And she'll come with a list, but I... | true | null | 2026-11-03T00:00:00 | ask-daniel | vocabulary | false | 2026-11-03: Bot built 2026-11-01 (Hebrish-Vocab-Telegram-Bot, submissions to a private dataset). Waiting on a BotFather token from Daniel to go live. planning/sprints/2026-11-03-snappy-advance-and-word-presets.md | null | null | null | 69.767 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T09:03:52.433Z |
20261101T090950Z_846pb9 | audio/20261101T090950Z_846pb9.flac | 2026-11-01T00:00:00 | 09:09:50 | Jewish and Talmudic vocabulary ideas | Capture app | [] | Daniel Rosehill | en | Yeah, just some random words I'm thinking of that we could use. Great rabbis, Ashkenazic and Sephardic, over time. Um, one honorific, zekher tzadik livrakha. Uh there's a few more like that. And great books, or sefarim. We've had a few which is nice, some of the major ones. Uh but there's still many, many more. Um and ... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | vocabulary | true | 2026-11-01: Pack religious-shas-sefarim (158 terms: masechtot, rabbis, sefarim, honorifics; 429 sentences), ASR-Training 4f7b5f8, release 124. | null | null | null | 61.183 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T09:09:50.371Z |
20261101T091055Z_yovt7t | audio/20261101T091055Z_yovt7t.flac | 2026-11-01T00:00:00 | 09:10:55 | Sourcing news vocabulary for speech data | Capture app | [] | Daniel Rosehill | en | I would say in general, the skill, I created already a float skill for getting words, but basically, it's finding a fertile source of vocabulary and noting it as a source in our internal saved resource. And using it to add in sentences. I mentioned the other day that I thought one way to make this more enjoyable for me... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | vocabulary | true | 2026-11-01: Pack israel-news-2026-11-01 (26 terms, 78 sentences); news sources and GDELT recorded in vocab-sources.json, ASR-Training 1f06e82, release 124. | null | null | null | 63.189 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T09:10:55.394Z |
20261101T092205Z_366nea | audio/20261101T092205Z_366nea.flac | 2026-11-01T00:00:00 | 09:22:05 | Background recording for capture app | Capture app | [] | Daniel Rosehill | en | All right. One thing I'd like to do is make sure that we have support for background recording in the capture app. So, what I'm thinking about doing is going on to Google Maps, zooming in on Jerusalem, zooming in on Israel, and I'll just like notch off a few of the areas that a few street names to add to the vocab list... | true | 2026-11-03T00:00:00 | 2026-11-03T00:00:00 | act | capture-app | true | 2026-11-03: Background recording shipped in Recorder release 123 (2026-11-01). | null | null | null | 51.503 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T09:22:05.668Z |
20261101T092305Z_m9xy6y | audio/20261101T092305Z_m9xy6y.flac | 2026-11-01T00:00:00 | 09:23:05 | Sync all button for upload queue | Capture app | [] | Daniel Rosehill | en | One other idea for the capture app is, I wanted to have the ability to like manually upload something, which, I can see it was added kind of, not exactly as I intended, like the upload queue every five is fine, I want to be able to upload that on demand. But maybe a more elegant way to do this would be simply to integr... | true | 2026-11-03T00:00:00 | 2026-11-03T00:00:00 | act | capture-app | true | 2026-11-03: Sync all (takes and ideas) shipped in Recorder release 123 (2026-11-01). | null | null | null | 35.439 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T09:23:05.999Z |
20261101T092818Z_k4s1vd | audio/20261101T092818Z_k4s1vd.flac | 2026-11-01T00:00:00 | 09:28:18 | Add Redback to vocabulary list | Vocabulary | [] | Daniel Rosehill | en | I'd like to add Redback to the vocab list. Redback is a type of boot, and suede boot. | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | vocabulary | true | 2026-11-01: Redback, pack gear-brands-2026-11-01, ASR-Training 2fff2ce; served as plain English (not code-switching) since release 124. | null | null | null | 8.943 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T09:28:18.004Z |
20261101T093041Z_8fl72e | audio/20261101T093041Z_8fl72e.flac | 2026-11-01T00:00:00 | 09:30:41 | Disabling general vocabulary recording by default | Capture app | [] | Daniel Rosehill | en | So, I think that there is actually no rationale in recording general vocabulary because I realized that on the the current training recipe we are using blending in some LibriSpeech. Therefore, I would say by default, it can be turned off, and only if I opt in, do I need to show general vocab, which I think we have alre... | true | 2026-11-03T00:00:00 | 2026-11-03T00:00:00 | act | capture-app | true | 2026-11-03: General English opt-in (off by default) shipped in Recorder release 123 (2026-11-01). | null | null | null | 26.343 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T09:30:41.468Z |
20261101T093139Z_dbifn4 | audio/20261101T093139Z_dbifn4.flac | 2026-11-01T00:00:00 | 09:31:39 | Niche vocabulary for speech recognition project | Vocabulary | [] | Daniel Rosehill | en | Okay, some other niche vocab to add is in terms of things I like and that transcription usually gets wrong, ToughBuilt, it's a company that make a specific type of belt with ClipTech as their technology, and then MOLLE as well, M O L L E, as in the US military attachment system. | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | vocabulary | true | 2026-11-01: ToughBuilt, ClipTech, MOLLE, pack gear-brands-2026-11-01, ASR-Training 2fff2ce; plain English since release 124. | null | null | null | 21.626 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T09:31:39.779Z |
20261101T093602Z_ovkgmd | audio/20261101T093602Z_ovkgmd.flac | 2026-11-01T00:00:00 | 09:36:02 | Shabbat and Jewish holiday vocabulary suggestions | Vocabulary | [] | Daniel Rosehill | en | So some vocabulary for Shabbat. Shabbat and Jewish law about Shabbat, Melakha, religious work, Mitzvot, Mitzvah, I feel like we could use a few more of those. And then Shabbat songs, Shalom Aleichem, Eshet Chayil, songs like that which in the, placing them in the appropriate context, Havdalah, Jewish holidays as well. ... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | vocabulary | true | 2026-11-01: Pack shabbat-calendar (57 terms, every Hebrew month at 3+, minor fasts, zemirot, minhagim; 155 sentences), ASR-Training 64d899a, release 124. | null | null | null | 54.41 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T09:36:02.323Z |
20261101T094008Z_6iob66 | audio/20261101T094008Z_6iob66.flac | 2026-11-01T00:00:00 | 09:40:08 | Add Seroquel to speech recognition data | Vocabulary | [] | Daniel Rosehill | en | I'd like to add Seroquel. It is an it is a medication, an antipsychotic, um and sentences should be in the context of sleep. It's used as a sleep medi | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | vocabulary | true | 2026-11-01: Seroquel and quetiapine in a sleep context, pack health-sleep-sharap-2026-11-01 (15 sentences with Sharap), ASR-Training a40ea26, release 124. | null | null | null | 14.367 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T09:40:08.307Z |
20261101T094034Z_zrg3cv | audio/20261101T094034Z_zrg3cv.flac | 2026-11-01T00:00:00 | 09:40:34 | Vocabulary note for speech recognition project | Vocabulary | [] | Daniel Rosehill | en | Uh one important vocabulary word is Sharap. It refers to a private medical service uh offered at Israeli hospitals. Um and all these would have like a few mentions in terms of uh just to get the vocab. Uh there might be another acronym for | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | vocabulary | true | 2026-11-01: Sharap, pack health-sleep-sharap-2026-11-01, ASR-Training a40ea26, release 124. | null | null | null | 16.759 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T09:40:34.942Z |
20261101T094117Z_1ch4r8 | audio/20261101T094117Z_1ch4r8.flac | 2026-11-01T00:00:00 | 09:41:17 | Adding Hebrew acronyms to speech recognition vocabulary | Vocabulary | [] | Daniel Rosehill | en | One particular type of vocabulary that we might wish to have is kitzurs. There's a Hebrew website called Kitzur It's a kitzur lookup, and it contains all the many, many acronym abbreviations used in Hebrew. And some of them might be sufficiently useful for English speakers in code-switching contexts that they could be ... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | vocabulary | true | 2026-11-01: 29 Hebrew acronyms checked on kizur.co.il, 65 code-switching sentences, pack hebrew-acronyms-2026-11-01, ASR-Training a40ea26, release 124. | null | null | null | 40.255 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T09:41:17.466Z |
20261101T094644Z_k2i30k | audio/20261101T094644Z_k2i30k.flac | 2026-11-01T00:00:00 | 09:46:44 | Vocabulary to integrate for transcription | Vocabulary | [] | Daniel Rosehill | en | Okay, so some vocab to integrate, as I'm pacing around the place looking and seeing what, you know, what are some things that that transcription gets wrong for me. So, in no particular order, Bosch, as in the tool company, Dremel, as in the both the tool and its use as a noun, so "I'm using my Dremel for engraving," fo... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | vocabulary | true | 2026-11-01: Bosch, Dremel, Ulanzi, Stanley, plug types, Loop and others: 26 terms, 70 sentences, pack tools-gear-brands-2026-11-01, ASR-Training a40ea26, release 124. | null | null | null | 121.989 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T09:46:44.861Z |
20261101T094921Z_r7gm86 | audio/20261101T094921Z_r7gm86.flac | 2026-11-01T00:00:00 | 09:49:21 | Adding Linux vocabulary to speech model | Vocabulary | [] | Daniel Rosehill | en | One other batch of vocab would be to go through my Ubuntu computer and see what I have installed in terms of CLIs, programs, and then see which ones are missing on the list. Ideally, trying to find the more popular tools that are, you know, very common on KDE Plasma, as well as the other major DEs on Linux. And then pr... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | vocabulary | true | 2026-11-01: 77 Linux tools from the laptop inventory and common DE apps, 212 tech sentences, pack linux-tools-2026-11-01, ASR-Training a40ea26, release 124. Plain-English packs no longer tagged as code-switching. | null | null | null | 23.973 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T09:49:21.606Z |
20261101T095605Z_99t4wl | audio/20261101T095605Z_99t4wl.flac | 2026-11-01T00:00:00 | 09:56:05 | Adding GPU and cloud vocabularies | Vocabulary | [] | Daniel Rosehill | en | One group to add to the AI vocab would be common GPUs. So let's say the most 10 or 20 of the most in-demand GPUs at the moment that would be available on serverless inference platforms. As well as runpod, beam cloud, Modal we have, any of the big Nebius, any of the big serverless and AI cloud platforms that are still m... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | vocabulary | true | 2026-11-01: Pack gpus-ai-clouds-2026-11-01: 33 GPUs and AI cloud platforms, 38 tech sentences, ASR-Training a40ea26, capture app release 124. | null | null | null | 32.127 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T09:56:05.881Z |
20261101T101709Z_fx7xs7 | audio/20261101T101709Z_fx7xs7.flac | 2026-11-01T00:00:00 | 10:17:09 | Adding makhzan to project vocabulary | Vocabulary | [] | Daniel Rosehill | en | One important vocab to add is makhzan. It means a warehouse and the context used is a bit more casual than warehouse would suggest. People often talk of having a makhzan in their house. I think we've had it a few times, but maybe not in the right context. It can also mean it can mean makhzan as in a professional wareho... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | vocabulary | true | 2026-11-01: 2026-11-01: machsan (canonical spelling) had 6 storage-room sentences; pack machsan-storage-room-2026-11-01 added 14 more in the everyday sense, the first pack built from a note's meaning, context and example phrasings (now pack fields). Generated on the OpenRouter Batch API. ASR-Training fdf4891, Recorder ... | null | null | null | 35.279 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T10:17:09.385Z |
20261101T102422Z_sidwpq | audio/20261101T102422Z_sidwpq.flac | 2026-11-01T00:00:00 | 10:24:22 | Adding Wolfson to vocabulary list | Vocabulary | [] | Daniel Rosehill | en | One new vocab to add is Wolfson and the context is this is a building in Jerusalem, very much populated with English speakers and context could be, I'm visiting my grandpa, my Safta in Wolfson. There's also a medical clinic there. I have an appointment coming up at the clinic in Wolfson. W O L F S O N. | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | vocabulary | true | 2026-11-01: Wolfson top-up (visiting Safta, the clinic), pack tv-channels-wolfson-maxstock-2026-11-01, ASR-Training 3517a47, release 124. | null | null | null | 27.354 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T10:24:22.892Z |
20261101T102528Z_i7jnnp | audio/20261101T102528Z_i7jnnp.flac | 2026-11-01T00:00:00 | 10:25:28 | Israeli TV channel speech data | Vocabulary | [] | Daniel Rosehill | en | One more is for Israeli TV channels. So Kan is station 11, which in Hebrew is arutz echad asar. And then you have channel 12, 13, 14. The most commonly referenced ones would be 12 and 13 by the arutz formulation, as well as by their informal channel names. 12 and 13 are more centrist, 11 or Kan is the national broadcas... | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | vocabulary | true | 2026-11-01: Arutz Echad, Shteim, Shlosh and Arba Esre (spoken forms), pack tv-channels-wolfson-maxstock-2026-11-01, ASR-Training 3517a47, release 124. | null | null | null | 47.087 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T10:25:28.853Z |
20261101T102649Z_sbtane | audio/20261101T102649Z_sbtane.flac | 2026-11-01T00:00:00 | 10:26:49 | Max Stock vocabulary for speech recognition | Vocabulary | [] | Daniel Rosehill | en | One more crucial piece of vocab to make sure we've enough coverage of is Max Stock. This is a discount chain throughout Israel, very popular for just buying random bits and pieces. Um so, we should have a few sentences in context about buying stuff at Max Stock, going to Max Stock. It's | true | 2026-11-01T00:00:00 | 2026-11-01T00:00:00 | act | vocabulary | true | 2026-11-01: Max Stock top-up, pack tv-channels-wolfson-maxstock-2026-11-01, ASR-Training 3517a47, release 124. | null | null | null | 16.09 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T10:26:49.989Z |
20261101T174041Z_nli12r | audio/20261101T174041Z_nli12r.flac | 2026-11-01T00:00:00 | 17:40:41 | Jerusalem vocabulary to add | Vocabulary | [] | Daniel Rosehill | en | Vocab to add, a steakia, which is a steak grill restaurant. The context would be we're going to eat as at a steakia. And I was just looking at Google News, and Givat Ram, neighborhood in Jerusalem was the Knesset. And I think there's actually quite a lot of neighborhoods in Jerusalem that we don't have yes. Beit Safafa... | true | 2026-11-03T00:00:00 | 2026-11-03T00:00:00 | act | vocabulary | true | 2026-11-03: Givat Ram, Har Gilo, Beit Safafa, Misrad HaChutz already in the bank. Added: Kniset HaIr, Givat Ram, Misrad Rosh HaMemshala, light rail Red/Green/Blue Line, steakiya: priority list, 10 sentences each. Skyscraper names not added (need names). planning/sprints/2026-11-03-snappy-advance-and-word-presets.md | null | null | null | 89.967 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T17:40:41.110Z |
20261101T174536Z_4vvwkb | audio/20261101T174536Z_4vvwkb.flac | 2026-11-01T00:00:00 | 17:45:36 | Local store vocabulary for speech project | Vocabulary | [] | Daniel Rosehill | en | there is a chains. New deli, really important, New Deli is a sandwich place. There is Lametayel, which is a hiking store. There is a lotto stand which is more generic type of business. There is Dr. Gav, which is like an ergonomic company for office chairs. There is Kravitz, which is another big office supplier. And so,... | true | 2026-11-03T00:00:00 | 2026-11-03T00:00:00 | act | vocabulary | true | 2026-11-03: Added New Deli, Lametayel, Dr. Gav, Tzidkiyahu, lotto stand, steakiya(ot), shipudiya, arais, entrecote, kasher lemehadrin: priority list, 10 sentences each. Kravitz, levivot, pargiot already covered. planning/sprints/2026-11-03-snappy-advance-and-word-presets.md | null | null | null | 111.151 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-01T17:45:36.683Z |
20261102T154801Z_bprlqx | audio/20261102T154801Z_bprlqx.flac | 2026-11-02T00:00:00 | 15:48:01 | Adding background noise to speech dataset | Annotation and training | [] | Daniel Rosehill | en | Here is a quick idea uh that I had. So I recorded some um noise floors, background noise in various environments ranging from the shuk to Yaffo street. Like 30 seconds sorry, 30 seconds, 1 minute up to 3 minutes in raw um unprocessed on my phone and I can get higher quality recordings on uh on a proper recorder device,... | true | 2026-11-03T00:00:00 | 2026-11-03T00:00:00 | backlog | training | false | 2026-11-03: Noise augmentation: ASR-Training-Set-Ideas noise-augmented-subset. Mics Samson Q2U, Zoom H1n, AT2020 added to the priority list (10 sentences each). | null | null | null | 84.282 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-02T15:48:01.804Z |
20261103T064713Z_kwth92 | audio/20261103T064713Z_kwth92.flac | 2026-11-03T00:00:00 | 06:47:13 | Predefined categories for vocabulary generator | Annotation and training | [] | Daniel Rosehill | en | Okay, so I'd like to be able to add from my vocab generator, so for example, there's a list of Jewish rabbis and streets in Jerusalem. So the word thing where I can like donate where I can, sorry, type a word and get sentences in the app, what I'd like to have is pre like I can do I can do custom or I can do it like a ... | true | 2026-11-03T00:00:00 | 2026-11-03T00:00:00 | act | capture-app | true | 2026-11-03: Word mode preset contexts (Street, Neighborhood, City in Israel, Rabbi, Jewish book, Synagogue, Israeli business, food, organization): Recorder release 151. planning/sprints/2026-11-03-snappy-advance-and-word-presets.md | null | null | null | 83.546 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-03T06:47:13.333Z |
20261103T064835Z_ly3kuy | audio/20261103T064835Z_ly3kuy.flac | 2026-11-03T00:00:00 | 06:48:35 | Multiple word input formatting | Annotation and training | [] | Daniel Rosehill | en | To build on my last idea, I think it would actually be most efficient even if I could do like multiple words either comma-separated or one per line. So what I'm envisioning is to come back | true | 2026-11-03T00:00:00 | 2026-11-03T00:00:00 | act | capture-app | true | 2026-11-03: Word mode takes several words, comma-separated or one per line, for sentences and priority: Recorder release 151. planning/sprints/2026-11-03-snappy-advance-and-word-presets.md | null | null | null | 74.437 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-03T06:48:35.063Z |
20261104T140148Z_6dwwfc | audio/20261104T140148Z_6dwwfc.flac | 2026-11-04T00:00:00 | 14:01:48 | Testing USB microphones for speech datasets | Annotation and training | [] | Daniel Rosehill | en | Yeah, just a quick thought. So, firstly, I am using this Oh, this is a good sample audio reference because I'm using my uh Jabra Evolve I keep forgetting the exact name. Jabra Evolve 30 2 It's exciting first uh milestone. I've never used a USB peripheral on my on an Android before. So, works. Um a little bit flaky. I w... | true | 2026-11-05T00:00:00 | 2026-11-05T00:00:00 | backlog | dataset | false | 2026-11-05: USB vendor and product ids: already saved on Android since release 183. Cork speech-to-text set: idea cork-stt in the training set backlog. Same recording as 20261104T140154Z_4dosw9. | null | null | null | 1,019.807 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-04T14:01:48.033Z |
20261104T140154Z_4dosw9 | audio/20261104T140154Z_4dosw9.flac | 2026-11-04T00:00:00 | 14:01:54 | Testing USB microphones for speech datasets | Annotation and training | [] | Daniel Rosehill | en | Yeah, just a quick thought. So firstly, I am using this, or this is a good sample audio reference because I'm using my Jabra Evolve, I keep forgetting the exact name, Jabra Evolve 32. An exciting first milestone, I've never used a USB peripheral on my on an Android before. So, it works. A little bit flaky, I would love... | true | 2026-11-05T00:00:00 | 2026-11-05T00:00:00 | drop | other | false | 2026-11-05: Duplicate of 20261104T140148Z_6dwwfc, saved 6 seconds later with a near-identical transcript. | null | null | null | 1,019.807 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-04T14:01:54.909Z |
20261104T150604Z_rueeh8 | audio/20261104T150604Z_rueeh8.flac | 2026-11-04T00:00:00 | 15:06:04 | Voice flux in fine tuning | Other | [] | Daniel Rosehill | en | just recording a quick note about the recording project. So, I guess the interesting thing about doing fine tuning of your voice is that voices are always in a state of flux. One ages. One wakes up with a terrible hangover and other things of this nature. And lately my wife and I have been coming | true | null | 2026-11-05T00:00:00 | ask-daniel | capture-app | false | 2026-11-05: Audio goes to digital silence at 27 s of 83 s, so the rest was never recorded: planning/bugs/to-resolve/BUG-0002.md. Waiting on Daniel to say the rest of the note again. | null | null | null | 83.293 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-04T15:06:04.935Z |
20261104T151246Z_hn8sdr | audio/20261104T151246Z_hn8sdr.flac | 2026-11-04T00:00:00 | 15:12:46 | Voice recognition issue with Jerusalem street name | Other | [] | Daniel Rosehill | en | Here is one more, a couple of more notes as uh you know as I'm walking occurred to me. So I was trying to put there is a particular street in Jerusalem that evades uh geolocations. HaKeren HaKayemet Le'Israel, the Jewish National Fund and it's not like searching an obscure term or street. But every single time, I try t... | true | 2026-11-04T00:00:00 | 2026-11-04T00:00:00 | act | vocabulary | true | 2026-11-04: Added to the priority list, high: HaKeren HaKayemet LeYisrael, Kakal (6 sentences each). Stored transcript stops early; the audio runs 94 s. | null | null | null | 94.026 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-04T15:12:46.273Z |
20261104T151700Z_tidyju | audio/20261104T151700Z_tidyju.flac | 2026-11-04T00:00:00 | 15:17:00 | Hebrew abbreviations in speech recognition | Other | [] | Daniel Rosehill | en | So, I'm still on my adventure picking up my my stuff here, and I just a couple of vocab words. I'm standing outside a laundry, it said mikhbasa, that's a laundromat. And then, there's one word that just occurred to me, it is KKL. So, Hebrew is full of these um little things that are called kitzurim, which are abbreviat... | true | 2026-11-04T00:00:00 | 2026-11-04T00:00:00 | act | vocabulary | true | 2026-11-04: Added to the priority list, high: mikhbasa, Kakal, kitzurim, Keren Kayemet LeYisrael (6 sentences each). | null | null | null | 121.551 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-04T15:17:00.938Z |
20261104T152452Z_flytw5 | audio/20261104T152452Z_flytw5.flac | 2026-11-04T00:00:00 | 15:24:52 | Bluetooth microphone quality for training speech recognition | Other | [] | Daniel Rosehill | en | All right, so I'm uh out and about at the moment. doing some errands and one more note from the field is the following. So, I have these Bluetooth headphones. I got them in the summer. They're called Soundcore P20i. Uh small p 20, small i. So, these are like Our dear little one, Ezra, uh oh my god, how many how many he... | true | null | 2026-11-05T00:00:00 | act | training | false | 2026-11-05: Noisy take A/B on a tiny set: planning/features/to-do/FEAT-0011.md, open question 59. Not run: needs a spend approval. | null | null | null | 402.287 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-04T15:24:52.271Z |
20261104T152808Z_jy6x34 | audio/20261104T152808Z_jy6x34.flac | 2026-11-04T00:00:00 | 15:28:08 | Training speech models with background silence | Other | [] | Daniel Rosehill | en | I've one more idea for training. So, I mentioned in these notes that I'm currently walking through Jerusalem and like a madman holding the holding my phone to get some training data. I am dedicated to the cause. And what I was thinking is What I was thinking about like silencing and I had a question to Claude regarding... | true | null | 2026-11-05T00:00:00 | act | training | false | 2026-11-05: Deliberate silence around sentences: planning/features/to-do/FEAT-0012.md, open question 60. Saved transcript stops early; the re-transcribed end is in the planning item. | null | null | null | 119.77 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-04T15:28:08.194Z |
20261105T104518Z_rd9vv5 | audio/20261105T104518Z_rd9vv5.flac | 2026-11-05T00:00:00 | 10:45:18 | Automated podcast creation from recorded updates | Other | [] | Daniel Rosehill | en | Here's one idea I have. I recorded an episode for the for the podcast, which is my longer form notes on how things are going, and firstly I hit a bug in the recording app, which we should look into and plot. And I thought we could have like category podcast, which would be we could have a data set. This is the idea. A ... | true | null | 2026-11-05T00:00:00 | act | website | false | 2026-11-05: Podcast category and feed: planning/features/to-do/FEAT-0010.md, a separate Hebrish task. Lost 10 minute recording: planning/bugs/to-resolve/BUG-0003.md. | null | null | null | 121.578 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-05T10:45:18.694Z |
20261105T104656Z_pi1ufs | audio/20261105T104656Z_pi1ufs.flac | 2026-11-05T00:00:00 | 10:46:56 | Vocabulary request for Israeli public companies | Vocabulary | [] | Daniel Rosehill | en | Okay, vocabulary request. Publicly traded Israeli companies, so two groups. The first would be companies listed on the Tel Aviv Stock Exchange. The objective here is to add to the lexicon companies beyond like let's say, Wix, plus plus the other kind of ones that are almost cliche, everyone's heard about them, because ... | true | 2026-11-05T00:00:00 | 2026-11-05T00:00:00 | act | vocabulary | true | 2026-11-05: Tel Aviv listed and foreign listed Israeli companies added to the priority list, 10 sentences each, as part of a 331-term run. Terms with 8 or more sentences skipped. | null | null | null | 73.557 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-05T10:46:56.377Z |
20261105T104958Z_orzke3 | audio/20261105T104958Z_orzke3.flac | 2026-11-05T00:00:00 | 10:49:58 | Adding urban planning vocabulary to app | Vocabulary | [] | Daniel Rosehill | en | Another type of vocabulary I would be very eager to include in the app. So, my wife, Hanna, is an architect and she is very passionate about urbanism and urban planning and I think feels that a lot of what is being done in Israel is like really terrible design. So, she uses a bunch of terms that I think people who feel... | true | null | 2026-11-05T00:00:00 | ask-daniel | vocabulary | false | 2026-11-05: Urbanism terms, thinkers (Jane Jacobs, taken as what 'Jane Austen' meant), Israeli planning terms, architects, mikhlala, mikhlalot and the colleges added to the priority list. Waiting on one term: 'line lies' in the note, to ask Hannah. | null | null | null | 155.039 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-05T10:49:58.360Z |
20261105T105208Z_7tkxdn | audio/20261105T105208Z_7tkxdn.flac | 2026-11-05T00:00:00 | 10:52:08 | Israeli place names for speech recognition vocabulary | Vocabulary | [] | Daniel Rosehill | en | So, one of the geographical, uh, place name things that I want to make sure we have in this is referring to place names in Israel is towns in Gush Etzion, which is the block of communities/settlements depending on your political orientation outside of Jerusalem. And we have the ring neighborhoods, but those are I think... | true | 2026-11-05T00:00:00 | 2026-11-05T00:00:00 | act | vocabulary | true | 2026-11-05: Gush Etzion, second tier West Bank towns, E1, Jerusalem Hills and Beit Shemesh area communities added to the priority list, 10 sentences each, neutral wording. | null | null | null | 118.49 | 48,000 | 24 | flac | capture-app | google/gemini-3.8-flash | 2026-11-05T10:52:08.027Z |
Hebrish Improvement Notes
Voice notes of ideas and incremental-improvement notes for the Hebrish fine-tune project and its recording app, captured in the moment and transcribed. Not training data: this dataset records what to improve and why, so the thinking behind the datasets, the annotation design and the training plans is kept in the open alongside the models.
Each row is one voice note: the original audio (audio/*.mp3) and its transcript, with a short
title, a category and tags. Later notes may supersede earlier ones — these are iterations.
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.
Columns
| Column | Meaning |
|---|---|
id |
stable id (YYYY-MM-DD-HHMMSS, the recording time) |
file_name |
the audio file (MP3 for the first three notes, FLAC 48 kHz / 24-bit for notes recorded in the app) |
date, time |
when the note was recorded |
title |
short descriptive title |
category |
grouped subject: Dataset design · Annotation and training · Technical ideas |
tags |
free tags for the specific ideas in the note |
speaker, language |
speaker and spoken language of the recording |
transcript |
cleaned transcript (fillers and false starts removed, nothing paraphrased) |
for_claude |
true when Daniel flagged the note for Claude (Claude Code) to act on |
claude_done_at |
date Claude acted on a flagged note, else empty |
Notes so far
| Date | Title | Category |
|---|---|---|
| 2026-10-26 | The Ideas dataset (meta-idea) | Dataset design |
| 2026-10-26 | Annotation design and the training plan | Annotation and training |
| 2026-10-26 | Technical ideas to explore | Technical ideas |
Speaker: Daniel Rosehill, the dataset's author, who speaks English with an Irish accent and lives in Jerusalem. Every recording is his voice.
How this dataset is meant to grow
The plan (described in the first note) is for the recording app used to collect the training data to gain an Ideas page: voice notes recorded there are pushed to this dataset rather than into the training sets, uploaded by a service worker, and transcribed with a Gemini Flash model (OpenRouter), with the transcript written back into the row. Until that exists, notes are added by hand.
Written by Claude.
- Downloads last month
- 725