Hacker News Reader

kenstleron SWE-Bench Pro

One of the authors here -- we should clarify that strong copyleft license is a best attempt at decontamination for the public set. It's part of the tradeoff for having an open source set -- true decontamination is available with the private commercial set, but we can't release these, and if we did they'd be immediately susceptible to future contamination.

kenstleron Show HN: NoteTech – Create personal automations by writing notes

Yes!

kenstleron Show HN: NoteTech – Create personal automations by writing notes

Presently our focus is more on enabling automations that integrate with external services, but we do intend to enable some phone actions that make sense in that context.

kenstleron Show HN: NoteTech – Create personal automations by writing notes

1. We can do one for web. 2. Would love to chat about it.

kenstleron Show HN: NoteTech – Create personal automations by writing notes

Love this - thank you!

kenstleron Show HN: NoteTech – Create personal automations by writing notes

Thanks for the feedback!

Yes - at present, a note needs to be deployed before you can test. Then you can make a change, and redeploy. We can look at making this a bit more seamless in future iterations.

Re: the AI button - any suggestion on how we might improve there?

kenstleron Show HN: NoteTech – Create personal automations by writing notes

Ah - I assumed you were talking about the in-app experience. My bad!

Point well taken and we'll definitely add more details/docs on the site.

Thanks!

kenstleron Show HN: NoteTech – Create personal automations by writing notes

Hi! A good general point.

We're doing a few things to try to guide the user:

1. We include ~26 fully-formed "Example" and "Template" notes in the app - these are intended to show users what they can do, but they can also be copied and used as-is or edited.

2. When users write new notes, we use a custom keyboard tool called "AutoPrompt" to show the user what actions they can take in each section. It includes prompts like "Search reddit for...", etc. that is inserted into the text entry and the user completes the statement. (You can see this in the vid on our home page).

3. If users write incorrect or wrong things in the note, they get a response back that instructs them what they need to correct.

Admittedly, as a notes-based interface there's a tradeoff we're trying to balance on flexibility vis-a-vis unstructured text input, and guardrails to inform the user. Appreciate any further feedback you can share!

kenstleron Show HN: NoteTech – Create personal automations by writing notes

Appreciate the feedback!

kenstleron Show HN: NoteTech – Create personal automations by writing notes

Yes, but instead of "compiling" directly to generated code, we generate an assembly of code modules.