Hacker Newsnew | past | comments | ask | show | jobs | submit | CharlesW's commentslogin

This isn't a good test for any model since LLMs can't math (even though frontier models can sometimes correctly simulate mathing), which is why one would always use a tool for this.

Oh! Thanks for that, I hadn't realized Ice's maintainer had stopped working on it: https://github.com/stonerl/Thaw

I had been using Hidden Bar but gonna give Thaw a shot now. Looks a bit more full featured and under active development!

https://github.com/dwarvesf/hidden


I personally found it confusing and un-Mac-like that quitting the configuration app also now stops the Tailscale service. It was unfortunate to discover this while I was AFK.

My recommendation is to rethink it to work like apps like 1Password, Default Folder, Keyboard Maestro, Ice, etc., where I can always easily open a configuration app, but the service must be intentionally/knowingly quit via either the configuration app or the menu bar utility.

TLDR: Please separate the service from the new configuration app.


This was driving me nuts but at some point very recently they added a "Hide Dock Icon" to settings.

> Its a CMS, designed from scratch…

Maybe not scratch scratch: "And under the hood, EmDash is powered by Astro…"

> It's built on top of, and seems to perfer you use CF proprietary capabilities (D1 Databases, R2 for image/media storage, their workers for running things.

D1 is SQLite, R2 is S3, and there are other ways to securely run plugins. If it was designed to only be possible to deploy on Cloudflare, they didn't do a very good job.


I'm building a commercial SaaS product on Workers. Although I've barely scratched the surface of what Cloudflare offers¹, so far it's been great. The value proposition is effectively the same as serverless in general: You worry about the product, they worry about deployment. Note that Cloudflare Workers is just one (albeit important) star in their constellation of capabilities.

¹https://developers.cloudflare.com/directory/?product-group=D...


This seems a great solution, and I'll definitely be trying it. I feel like monospace fonts are the Roman roads → horse ruts → rail gauge of our industry.

The title and of this article is Don't Let AI Write For You, when its point seems to be closer to Don't Let AI Think For You (see "Thinking").

This distinction is important, because (1) writing is not the only way to faciliate thinking, and (2) writing is not neccessarily even the best way to facilitate thinking. It's definitely not the best way (a) for everyone, (b) in every situation.

Audio can be a great way to capture ideas and thought processes. Rod Serling wrote predominantly through dictation. Mark Twain wrote most of his of his autobiography by dictation. Mark Duplass on The Talking Draft Method (1m): https://www.youtube.com/watch?v=UsV-3wel7k4

This can work especially well for people who are distracted by form and "writing correctly" too early in the process, for people who are intimidated by blank pages, for non-neurotypical people, etc. Self-recording is a great way to set all of those artifacts of the medium aside and capture what you want to say.

From there, you can (and should) leverage AI for transcripts, light transcript cleanups, grammar checks, etc.


> Audio can be a great way to capture ideas and thought processes ... This can work especially well for people who are distracted by form and "writing correctly" too early in the process, for people who are intimidated by blank pages, for non-neurotypical people, etc. Self-recording is a great way to set all of those artifacts of the medium aside and capture what you want to say.

Yes, this is my process:

Record yourself rambling out loud, and import the audio in NotebookLM.

Then use this system prompt in NotebookLM chat:

> Write in my style, with my voice, in first person. Answer questions in my own words, using quotes from my recordings. You can combine multiple quotes. Edit the quotes for length and clarity. Fix speech disfluencies and remove filler words. Do not put quotation marks around the quotes. Do not use an ellipsis to indicate omitted words in quotes.

Then chat with "yourself." The replies will match your style and will be source-grounded. In fact, the replies automatically get footnotes pointing to specific quotes in your raw transcripts.

This workflow may not save me time, but it helps me get started, or get unstuck. It helps me stop procrastinating and manage my emotions. I consider it assistive technology for ADHD.


Writing is, however, a uniquely distinct and well-studied way to facilitate thinking.

I've definitely lost something since migrating my Artist's Way morning pages and to the netbook. (Worth it, though, to enable grep—and, now, RAG).


Yeah this is my problem. I can come up with ideas, but in writing my ideas never come out well. AI has helped me to express my ideas better. People who write well or are successful at writing sometimes fail to understand how uncommon is it to actually be good at writing. Shit is hard.

I would count direct dictation (eg someone writes down what you say, and that is the final text), as writing, in the context of producing a document (book, etc) that you intend others to read.

It's not the same thing as talking to someone (or a group) about something.


I'm finding AI great to have a conversation with to flesh out ideas, with the added benefit it can summarize everything at the end

You're being steered without being aware of it.

Worse. You’re being steered along a circle

Maybe they are aware of it?

I talk to other people. They influence me, steer me. I am okay with that.


not at all, it's very productive.

I do this a lot. Start by telling the AI to just listen and only provide feedback when asked. Lay out your current line of thinking conversationally. Periodically ask the AI to summarize/organize your thoughts "so far". Tactically ask for research into a decision or topic you aren't sure about and then make a decision inline.

Then once I feel like I have addressed all the areas, I ask for a "critical" review, which usually pokes holes in something that I need to fix. Finally have the AI draft up a document (Though you have to generally tell it to be as concise and clear as possible).


I usually create a document/folder with my thinking on what I want to do, any background information that is relevant, conversations on the topic, technical manuals, links etc. Then enter a conversation and explore the problem space and do something very similar to what you are doing.

I know you're not being serious, but for anyone who may not realize that, it does more than disabling attachments. Lockdown Mode's "optional, extreme" protection substantially changes the experience of using your device. https://support.apple.com/en-us/105120

How is this different and/or more interesting than Superpowers' episodic-memory skill¹ or Anthropic's Auto Dream²?

¹ https://github.com/obra/episodic-memory ² https://claudefa.st/blog/guide/mechanics/auto-dream



the biggest difference would be the /foresight

I built https://pwascore.com/ several months ago.

Where pwa.gripe cherry-picks and has an axe to grind, pwascore.com is intended to be a more thorough and dispassionate evaluation. I will add desktop browsers soon.

Click "Expand All" for a complete and detailed list. Click "How Scores Work" to understand the scoring heuristics.


Well the notifications on your web site doesn't provide the full idea of how limited it is in Safari. We have to ask users to install the shortcut before allowing notifications with Safari. Guess how many users actually go through with installing the shortcut? Nearly none.

If we had to ask users to go into their settings and switch the "enable notifications" flag we wouldn't call that supporting anything. The whole process of installing a shortcut to even get to the point where we can ask for notifications is even more convoluted on iOS.


How did you determine the weighting for your scores?

The most significant effect is that experimental and non-standard PWA capabilities aren't reflected in the primary score. You can see raw/unweighted scores by hovering over the primary score. Chrome wins handily if you count experimental/non-standard features.

For standards-based features I used a 4-tier model, described about halfway through the README (which I should also add to About):

    ┌────────┬──────────────┬───────────────────────────────────────────────┐
    │ Weight │ Tier         │ Rationale                                     │
    ├────────┼──────────────┼───────────────────────────────────────────────┤
    │ 3.0    │ Core PWA     │ Prerequisites for production PWAs (6 features │
    │        │              │ features: Web App Manifest, Service Workers,  │
    │        │              │ Caching, HTTPS, etc.)                         │
    ├────────┼──────────────┼───────────────────────────────────────────────┤
    │ 2.0    │ Important    │ Enhance PWA functionality (18 features: Push  │
    │        │              │ API, Add to Home Screen, Offline Support,     │
    │        │              │ Display Modes, etc.)                          │
    ├────────┼──────────────┼───────────────────────────────────────────────┤
    │ 1.0    │ Standard     │ Default weight (94 features)                  │
    ├────────┼──────────────┼───────────────────────────────────────────────┤
    │ 0.5    │ Experimental │ Nice-to-have capabilities (43 features:       │
    │        │              │ Sensor APIs, Bluetooth, NFC, AR/VR, etc.)     │
    └────────┴──────────────┴───────────────────────────────────────────────┘
This weighting turns out to be reasonably conservative. For example, if you hover over the score for Firefox (the largest benefactor), you'll see that it bumps Firefox's score by 5.

I'm very open to feedback. This is a sincere attempt to quantify vendors' PWA support.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: