We build the evaluation datasets and RL environments that make AI reliable where mistakes are expensive: finance, healthcare, and legal. We design expert-curated training data, calibrated rubrics, and RL environments for frontier AI labs and startups pushing the frontier of what models can do.

We're a small London-based team running multiple active projects, so what you ship gets used immediately by labs, startups and internal domain experts. We recently launched SpreadsheetBench v2 and work on ultra long-horizon tasks.

We're hiring:

* MTS, Applied AI — design the next benchmarks and RL environments — https://jobs.ashbyhq.com/aptura/c299e1f2-cdb4-4842-b5f3-8aa6...

* MTS, SWE / Product — build the platform behind every dataset — https://jobs.ashbyhq.com/aptura/cd6b6df2-1a75-442e-83b3-2ae0...

Comments URL: https://news.ycombinator.com/item?id=47067128

Points: 3

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=43935110

Points: 2

# Comments: 0

Thats why I made Deadlines: A widget app that syncs with Apple Calendar & Reminders and allows you to filter them by keyword (supports Regex). Staying always up to date without any extra work from your side.

Nowadays I just add #track or similar as a note to my important events and know that I won't overlook them.

Recently I just added the option to open events & reminders in any app of your choosing. Also supports custom URL Schemes and Shortcuts.

Let me know what you think :)

Comments URL: https://news.ycombinator.com/item?id=40523789

Points: 2

# Comments: 0

The app is like a customisable board book:

A realistic picture on the left page, then the word in 1 to 4 languages of your choice on the right page. When you tap on the picture, it changes and a realistic sound is played. The words can also be read out loud using text to speech.

I used Midjourney to create "photos" for each word.

- I chose this style because realistic images have been shown to help toddlers the most with real-world recognition. Detailed illustrations would have the same benefit, but it was too difficult to get a consistent style.

- Most photos were generated in a few tries. Generating a dinner knife, a potty and the correct number of objects was quite difficult. A single banana was impossible for me. Even with image prompts I could only create pairs.

Translations were done manually using dictionaries, and some languages were double-checked by native speakers.

- Arabic, Chinese, Hindi, Japanese and Korean also show their romanisation to help non-native readers.

Hope you too find the app useful.

Comments URL: https://news.ycombinator.com/item?id=38445312

Points: 1

# Comments: 0

The app works similar to Rules in Mail:

- Specify some conditions (e.g. Calendar is "Work", Location contains "zoom")

- Choose shortcuts to run on events that meet the conditions

     - you can have multiple actions, each with a different offset and custom input

- The app can only trigger automations while your Mac is awake (missed actions can be triggered on wake up)

- The free version offers full functionality, but is limited to a max of 2 rules. Pro is a one-time purchase

- All your data stays on device + no ads or data collection

I would appreciate any feedback, especially what automations you might use the app for

Comments URL: https://news.ycombinator.com/item?id=37158944

Points: 4

# Comments: 0

So it would be $900 per day

Comments URL: https://news.ycombinator.com/item?id=34986545

Points: 1

# Comments: 0

Sure if the repo has a xcodeproj you can just open it in XCode, change the signing and improve on it. (Just make sure to always respect the licenses) If you want to play with whisper.cpp you can use this SwiftUI demo: https://github.com/ggerganov/whisper.cpp/tree/master/example...

To get out of this mindset, I decided to make an app for myself in a week, add monetization, release it and move on. The app idea was simple: Reflect on your day by answering the same four questions out loud. The answers are transcribed and with regular use you can see what influences you the most and take action. All on-device, as otherwise I wouldn't feel comfortable sharing my thoughts.

I had all core features working within a day by simply modifying an existing example app. However I was dissatisfied with iOS's built-in offline transcription due to a lack of punctuation and the speech recognition permission prompt that made it seem like data would leave the device. Decided to use whisper.cpp [0] (small model) instead. This change, lead to many others, as I now felt too little of the app's code was mine. e.g.:

  - Added automatic mood analysis. First using sentiment analysis, then changed to a statistical approach
  - Show trends: First implemented TextRank to provide a summary for an individual day, then changed it to extract keywords to spot trends over weeks and months. Replaced TextRank with KeyBERT for speed and n-grams, then BERT-SQuAD, and ended on a modified YAKE [1] for subjectively better results. (Do you know of a better approach?)

  - Transcription is not live but performed on recordings, so if you immediately want the transcript of your most recent answer, you have to wait.
  - Mood and keyphrase extraction are optimized for my languages and way of speaking, so they might not generalize well.
  - Music in the background can result in nearly empty transcripts.

[0] Show HN: Whisper.cpp https://news.ycombinator.com/item?id=33877893

[1] YAKE: https://github.com/LIAAD/yake

Comments URL: https://news.ycombinator.com/item?id=34868771

Points: 6

# Comments: 3