A solution, in theory

One way to prevent your web content from being embedded on another site in a frame specifically is to include the X-Frame-Options header in the HTTP response, with the policy set to DENY or SAMEORIGIN. I prefer DENY because I don’t think I’ll ever need to embed my own blog content, but it seems like SAMEORIGIN is more common.

OK, simple enough, but how to make this happen?

A fly in the ointment

This blog is built with the Quarto framework and published on Netlify.

Let’s start with Netlify, because that’s there the HTTP response header is ultimately going to be served from. Netlify’s Custom Headers page says we just need to include a file _headers in our publish directory and even gives an example for the clickjacking problem:

/*
  X-Frame-Options: DENY
  X-XSS-Protection: 1; mode=block

OK, now back to Quarto, where we find the fly in the ointment.¹ We need to get this file into our publish directory (by default this is called _site). But we can’t just put the file into the _site directory because Quarto deletes it every time it renders the source.

The Quarto docs do have a page about headers, but they’re HTML headers, not HTTP, so that’s no help.

The solution, for real

Instead, we need to refer to Quarto’s site resources page, which shows how to include files like _headers (or other things like images) in the _site directory after rendering.

In the root directory of our project we have a config file called _quarto.yml. In the project field of that file we list _headers as a resource, which tells Quarto to copy the _headers file over every time it renders.

project:
  type: website
  resources:
   - "_headers"

So just leave the _headers file in the root directory. Render the changes and publish to Netlify and we’re good to go.

$ quarto render
$ quarto publish --no-render netlify

To verify, open the dev tools in your favorite browser and go to the network tab. Open (or refresh) your blog and click on the top entry, which should be the page itself (not a secondary resource). To the right, you should see a section for Response Headers, and you should see the new X-Frame-Options entry.

The power of cross-platform search

Let’s say, for example, we want to follow the new Star Trek TV show, so we open the Apricot app, go to the Add feed, and search for “Star Trek”. We see something like this:

Apricot searches all platforms for Star Trek, so in addition to the TV show results, we also see two Spotify podcasts. In fact, if it weren’t for the ongoing Reddit blackout, we would also see several related subreddits.

We can follow all of these feeds in Apricot, and geek out on as much Star Trek content as we can take. We just click the icon to the right of each feed we want to follow and watch the icon change to an orange check mark. It should look something like this:

Now we’ll see both the TV show episodes and podcast episodes in our Apricot feed—super cool!

Looking up the road

If you’re a Star Trek fan, you probably already noticed a problem: the upcoming show isn’t actually in our results list. That would be Star Trek: Strange New Worlds. Let’s add the word “strange” to our query:

Ah, there it is.

One of the weaknesses with Apricot’s first version of search is that the query has to be pretty close to the true feed title. Looking ahead, we’ll be working on fixing this with fuzzy searches that tolerate partial matches and typos better, plus some other biggies:

• A one-line description of each search result. Today, you can click on each result to see the feed in the original platform, but we know that’s not ideal.

• A bigger set of feeds to search from. The more feeds our users add to Apricot, the better the search results will be for everybody.

• AI-generated feed suggestions, when you don’t have a specific feed in mind.

• Other languages besides English. Apricot search works ok for non-English feed names, but it’s really optimized for English.

Try it out and let us know what you think! Stay tuned for more updates and another blog post on the technical details of our search implementation.

Some context

This is not a scientific study, clearly. Once upon a time, I enjoyed doing controlled, in-depth, technical comparisons of ML models, but those days are past. In this post, I’m going to take about an hour to explore a few use cases, make a decision, and move on to the rest of my long to-do list.

The competitors

• OpenAI ChatGPT, with the GPT-4 model, in the chat window at https://chat.openai.com/. I pay about $20/month for this.

• Google Bard, in the chat window at https://bard.google.com/. I’m not sure what the model is under the hood, so minus 1 point for Google. But it’s free, so plus 1 point and we’re back to even.

Challenge and Results

import xml.etree.ElementTree as ET

def parse_opml(file_path):
    # Parse the OPML file
    tree = ET.parse(file_path)
    root = tree.getroot()
    
    # OPML should have a 'body' child, which contains 'outline' elements
    body = root.find('body')

    outlines = []

    for outline in body.findall('outline'):
        # Each outline should have 'text' attribute
        text = outline.get('text', '')
        
        # Append it to our list
        outlines.append(text)

    return outlines

def parse_opml(filename):
  """Parses an OPML file.

  Args:
    filename: The path to the OPML file.

  Returns:
    A list of dictionaries, where each dictionary represents an outline item.
  """

  with open(filename, 'r') as f:
    xml = f.read()

  # Parse the XML into a tree.
  tree = ET.fromstring(xml)

  # Get the outline items.
  outline_items = []
  for outline_item in tree.findall('outline'):
    outline_item_dict = {}
    outline_item_dict['title'] = outline_item.find('title').text
    outline_item_dict['url'] = outline_item.find('xmlUrl').text
    outline_items.append(outline_item_dict)

  return outline_items

And the overall winner is….

ChatGPT.

ChatGPT beat Bard pretty convincingly in summarization and Python coding. That’s worth $20/month to me, at least for now, despite the fact that they tied in Tweet and marketing tactics generation.

I’m happy there’s competition and I suspect Bard will improve quickly—I look forward to re-doing this test again soon!

How to install Apricot

Here are the steps to install Apricot on iOS with Safari. Check out the video above for a visual demo of me going through the steps on my own phone.

1. Open Safari

2. Navigate to the Apricot app, either directly or via the landing page. No need to login (although it’s ok if you do).

3. Press the share button in the middle of the Safari control bar at the bottom of the screen. It looks like this:

   

4. Scroll down until you see the item that says Add to home screen

   

5. Press Add in the top right. And voilà, there’s the icon for the Apricot app, on our home screen.

Some caveats, I’m afraid

• You’ll have to log in again in the app version of Apricot, even if you’re already logged in in the Safari version.

• You’ll have to use a Google account to log in to the app version of Apricot; the email address option doesn’t work. I know it’s annoying, but I think it’s an inherent limitation of iOS and I don’t see any way around it.

Feedback welcome

Everybody’s set up is a little different, so don’t be afraid to reach out and let me know how it goes for you.

Send feedback to support@theapricot.io or fill out the feedback form.

How to install Apricot

Here are the steps to install Apricot on Android with Chrome. Check out the video above for a visual demo of me going through the steps on my own phone.

1. Open Chrome

2. Navigate to Apricot, either directly or via the landing page.

3. Click Chrome’s triple dot icon in the top right (I called this the “settings” menu in the video, which isn’t quite right, but I don’t know what else to call it).

4. About three quarters of the way down you’ll see an option to Install App. Press that.

5. In the window that pops up press Install to confirm your choice.

6. Wait a few seconds until the system confirms installation is complete.

7. Close the browser. The Apricot icon will appear on the home screen. Move it to wherever you like, just like any other app. Press to open Apricot. Voilà!

A couple notes

• You don’t need to be logged in to install Apricot to the home screen.

• Sometimes a pop-up will appear when you first open Apricot in Chrome that asks if you want to install it. This is in fact the easiest way to do it, but sometimes it doesn’t appear. If it doesn’t, that’s ok, just follow the video or instructions above.

• Every browser is different; let me know how installation goes for you, especially if you have trouble. Send feedback to support@theapricot.io or fill out the feedback form.

How to enable and disable images

Most content platforms—even podcasts—include an image with each item. I recently enabled Apricot to show these images in feeds because they often add a lot of information about individual items.

Many Reddit posts, for example, are just like “What is this bug?” with an attached photo. Without the image, you have to click and wait for the Reddit page (or app) to load, which takes several seconds. With the image in Apricot, you can decide instantly whether it’s worth your time to click and contribute to the thread.

The problem is that images take up a lot of screen real estate, especially on a phone, which makes it hard to scroll through many items quickly. If you have lots of feeds or feeds with many items, (e.g. HackerNews), it can feel like images just get in the way.

Fortunately, Apricot enables both browsing “modes”. Turn images on or off:

• Go to the settings page (click the icon in the nav tray)

• Look for the App Settings section at the top of the page.

• Check or un-check the box next to the “Show images” line.

That’s it! Happy browsing!

Use case 1: stay in the loop

The simplest and most common use case for Apricot is to scan what’s new across all your feeds, what I’ve called “staying in the loop” in a previous post. The goal is simply to see all items from all feeds, in one interleaved list.

This is the foundational purpose of Apricot so it gets a whole page to itself: click the icon in the nav bar. This is Apricot’s home page, in fact, so usually you’ll already be on this page when you open the app.

Use case 2: binge mode

Sometimes we want to know what the latest content is from one specific feed.

For me, this usually happens when I discover a new Substack newsletter or Spotify podcast and I want to binge that content without distractions from other feeds. It’s also really useful for educational feeds, even if you’re not bingeing. When it’s time for me to study Spanish, for example, I want to know if there are any new videos from the Easy Spanish YouTube channel.

For this use case, click the icon in the nav bar to see the list of your feeds, then click the name of the feed you want to browse.

Use case 3: contextual browsing

At certain times of the day, I want to find good content on a particular type of platform, but not limited to a specific feed.

When driving, for example, I want to listen to a Spotify podcast. After dinner, I want to watch a TV show. Sunday morning, I like to read long-form Substack newsletters over coffee.

To browse content from a specific platform, go to the feeds list again (), then click any platform. You’ll items from feeds from that platform, but nothing else.

How it works

How it doesn’t work

Following TV shows on Apricot doesn’t make sense for all use cases.

In general, Apricot’s goal is to help you stay in the loop with new content as it’s published. So it doesn’t really make sense to follow old TV shows on Apricot. You can, but it’s not going to be worth much.

Apricot also doesn’t help to alert you about episodes before they air, so it’s not great for network TV shows unless there’s an accompanying streaming version you can watch later (e.g. The Late Show with Stephen Colbert is also available streaming online).

And it’s of limited usefulness with shows on Netflix, which tends to dump whole seasons at once. Apricot will help you to know when the season drops, but that’s about it.

Let me know what you think!

I’m really excited about this feature, even though it has absolutely nothing to with AI. Let me know how it goes for you and what other features you’d like to see in Apricot. 🙌

-Brian

How to generate summaries

Any item in Apricot that has the    icon in the top right can be summarized. Click or press the icon to generate the summary. Click it again to make the summary go away. It’s easier to just show you—check out this quick video.

Why bother? Some examples

Let’s look at a couple examples to get a sense for how Apricot’s summarization works and how it helps. First, an example from the Hacker News front page.¹

The first item in this screenshot is Nearly all apps of generative AI are a net negative for society. This is a “hot-take” title whose goal is to persuade you to click by arousing strong emotion without revealing the true take-home message.²

Apricot’s summary shows the author’s true fear is “fake content”, which is a much narrower concern than what the title suggests. Don’t know about you, but I’m not gonna bother clicking that one.

The second item in this screenshot does not have the    icon, which means Apricot can’t generate a summary. This happens when an item simply doesn’t have enough content, which is common with link- or image-only posts on Reddit.

Here’s another screenshot from a Reddit feed, namely r/movies.

The second item—Girl With the Dragon Tattoo, did you notice….—is pure clickbait. Apricot’s summary tells us what the content is really about: the user discovered a hidden scream in the movie. I’m actually intrigued by that—more than I am by the author’s own title—so I’ll click and learn more.

Limitations

Summarization is super powerful but there are some many caveats.

Testing in prod

RSS feeds are total chaos compared to the predictable schema and formatting of Substack newsletters. In my internal testing, I’ve been happy with Apricot and ChatGPT’s ability to handle all sorts of inputs, but I can’t possibly cover all corner cases myself. So, please don’t hesitate to send feedback if you see anything that looks buggy. I’m particularly interested in the quality of Apricot’s summaries, especially for non-English content.

Export your feeds from Feedly

1. Open Feedly in a web browser and log in.

2. Go to the “Organize Feeds” page. Click the gear icon to the right of your feed list.

   

3. Go to the “Export OPML” page. This one is a bit hidden. Click the arrow icon on the top right of the page, next to the “Import OPML” button.

   

4. Click the big green DOWNLOAD YOUR FEEDLY OPML button and save the output file.

   

Import your feeds into Apricot

1. Open Apricot, at https://app.theapricot.io.

2. Log in or sign up using either a Google account or your email. No password needed.

3. Click the Settings icon in the nav bar to go to the settings page.

   

4. Click the button under “Import feeds from OPML file” and select the file you exported from Feedly.

   

5. Apricot will start working as soon as you select the file. It takes a few moments to fetch and validate each feed, so be patient. When Apricot is done you’ll see a pop-up window that lists any issues we found during the import.

Wrapping up

Once your feeds are imported, Apricot redirects to the Feeds page, just to confirm your feeds were imported successfully. From there, click on any individual feed to browse, or click the Items page to see items from all of your feeds interleaved in one place.

OPML import is a new feature for Apricot, so don’t hesitate to send feedback if you run into trouble or have suggestions.

Thanks for trying Apricot!

Staying in the loop

It’s one of those things that’s easy to spot but tough to define. ChatGPT defines staying in the loop as:

to remain informed or up-to-date about current events, developments, or ongoing conversations, typically in a specific field or context. It often implies actively following news, updates, or discussions to maintain awareness and not be [sic] left out of important information or decisions.

It’s tempting to equate it with geopolitical current events, but that’s not what I think most people are really interested in. For most of us, it’s more about things like:

• seeing the latest photos of the grand-kids,

• reading/watching/listening to trusted commentators digest news and trends,

• discovering hot new shows to watch on TV,

• keeping a finger on the pulse of trends in our profession, and

• seeing job posts, real estate listings, or for-sale ads as soon as they’re published.

Why do we feel such FOMO a strong desire to stay in the loop? That’s above my pay grade, but I suspect it mostly comes down to two reasons: participating in a shared social context and quickly identifying risks and opportunities.

What’s not in the loop

Because the idea of staying in the loop is fuzzy, it’s also helpful to delineate what’s not part of staying in the loop.

• Content itself

• Search

• Real-time communication

A brief history of staying in the loop

The tried and true method to keep up-to-date with things is, of course, talking directly to other people. It’s a fundamental part of our existence.

Many people who grew up before the internet stay in the loop with the morning newspaper, the evening news on network TV, and maybe a radio show on the drive home from the office. Each of these outlets is a single company with one editorial team, curating the items they think will matter to their audience across a wide range of subjects.

With the rise of the internet, we began to seek content from a broader variety of sources. RSS readers were born to help us keep track of our favorite sources and to avoid having to click all around the web. RSS readers require extra work from the reader, though, for little benefit. They quickly fell into a niche market.

Instead, it was social media that took the world by storm in the late 2000s, quickly becoming the dominant way to stay in the loop for almost everything. No need to belabor this point—almost everybody on the planet knows how pervasive social media is.

Surprisingly, the humble email newsletter also made a noteworthy comeback. Now, every content platform wants us to subscribe and follow not only on social media but in our inbox as well, so they can bombard us with messages they know we can’t ignore.

Staying in the loop about staying in the loop

I’m not alone in realizing the value of helping users stay in the loop. Spotify recently overhauled the home page of its app to emphasize discovery. In a great interview on the Decoder podcast, Spotify exec Gustav Söderström laid out the reasons for the change:

When we ask users what they love…what they really say is, “The more new things you show me, the more new things I find, the more I’m going to like this product”… Discovery is kind of the lifeblood of Spotify. If you take that lens, then you should expect us to try to do everything to improve discovery.

Substack also announced just this week that it’s going in a similar direction, adding a product called Notes. Their rationale:

…Notes is designed to drive discovery across Substack…Our goal is to foster conversations that inspire, enlighten, and entertain, while giving writers a powerful growth channel as these interactions find new audiences.

Reddit’s product updates in March also seemed oriented toward discovery, with new feeds specifically for reading and watching videos and a more streamlined interface to make it “easier for people to find relevant content”.

You get the point.

The next revolution

All of these efforts face an insurmountable obstacle: they only help you to discover content on their own platform. A key part of staying in the loop, though, is finding the relevant stuff across the web. And that’s exactly where Apricot comes in.

Apricot’s only mission is to help you stay in the loop. Follow content from a variety of text, audio, and video platforms, interleaved into one personalized feed. State-of-the-art AI prioritizes the most relevant content and summarizes content to help you see through clickbait headlines. And, coming soon, a Discover page of our own, to help you find things you didn’t even know you were looking for!

Check it out—discover Apricot at https://app.theapricot.io and stay in the loop!

Attention is all you need¹

The attention economy has even the most reputable publications resorting to clickbait headlines. Here’s a classic, from Adam Grant’s Substack newsletter, which I’ve subscribed to in Apricot:

If You Do This, Your Emails Might Be Rude

By now, you know not to give in immediately to the curiosity—do WHAT?—but unless you click, you still don’t know if it’s worth the time or not.

It’s not just clickbait, though. Many writers don’t bother with descriptive titles, especially for newsletters, which the authors know are going to your high-attention email inbox.

The other headlines in my Apricot screenshot belong more to this category:

The public imagination

Adam Sandler and then some

Scare the rich

With Apricot, you can subscribe directly to Substack newsletters and keep them out of your inbox. But then how do you know whether to open content like this?

Summarize with Apricot

Check out the    icon to the right of each item in the Apricot screenshot. When you click that button, Apricot shows a succinct, one-sentence summary of the article.

In this example, it looks like the Adam Grant article is pretty obvious stuff—don’t write cold emails that ask a stranger to jump on a call immediately, etc. I’m gonna skip that one, thanks. The Bill Murphy, Jr. piece, on the other hand, looks interesting. I don’t care too much about Adam Sandler, but an Italian orange-throwing festival? Now you’ve got my attention.

The Stuff You Should Know episode does not have the    icon next to it. For now, summarization is only enabled for Substack newsletters, but it should be enabled for the other platforms soon.

How it works

Magic ChatGPT.

As of today, Apricot uses the GPT-3.5-turbo model.

The input to the model is the concatenation of the article title, subtitle, and full text, truncated to fit within the model’s context length constraint. The prompt simply asks ChatGPT to summarize the text in one sentence. No fancy spells prompt engineering at this point.

What’s the catch?

I’m super excited about this new feature, but I’ll also be the first to admit that there are lots of rough edges.

• Most importantly, ChatGPT is not perfect. With complex articles, it can lose track of the main ideas. With very short articles, it might hallucinate. So don’t rely heavily on these summaries for high-stakes applications. Seriously, I can’t stress that enough. When in doubt, just click the item and read the source.

• Apricot does not currently have a way to get the full text of pay-walled articles. Sometimes there’s enough material in the title, subtitle, and teaser, but this is the kind of situation that can lead ChatGPT to hallucinate. To minimize this, Apricot will just decline to summarize if there’s not enough content.

• The ChatGPT API is slow. It usually takes several seconds, so please have patience. The good news is that each item only needs to be summarized once across all Apricot users; subsequent requests are just database queries for the stored summary, so they’re much faster.

All your feedback are belong to us!

I’m confident our new ChatGPT-powered summarization feature will change the way you browse content. As always, I’d love to hear your feedback, so please don’t hesitate to reach out and share your thoughts.

Happy summarizing!

1. Sign up

Sign up for Apricot with an email address or a Google account. No password is required for the former—you’ll get an email with a magic link that confirms your address and logs you in for the first time.

2. Add feeds

After you sign up, Apricot will send you directly to the Add Feeds page.

The first order of business is to add feeds to follow, just as you would add people to follow on a social media platform. The difference is that with Apricot, your subscriptions come from different platforms, namely Spotify, YouTube, Substack, and Reddit. You can also add RSS feeds.

To get started, add a few of the listed popular feeds (don’t worry, you can remove them later) by pressing the gray plus icon to the right. Later, once you’re settled in, you can come back and add other feeds by pasting their links into the top part of the page.

Got a couple of feeds added? Great, now click the icon in the nav bar to browse the content from your feeds.

3. Browse

Apricot’s Items page is where the magic happens—think of it like your main feed on Twitter, LinkedIn, or Facebook. Apricot automatically imports items from all the feeds you’ve subscribed to and lists them here, as they come in.

There are four things you can do on the Items page.

• Open an item by clicking the text of the item. In the example image above, let’s say we want to listen to the Stuff You Should Know podcast episode about semantic satiation. Clicking the text will open the Spotify app or website directly to that episode’s page.

• Bookmark an item by swiping right (on mobile) or clicking the bookmark icon (on desktop). Bookmarking is useful to save items permanently or temporarily, e.g. as a “to read” list. In the latter, case, swipe left to un-bookmark after you’ve read/listened to/watched the item.

• Go to a feed page by clicking the title of an item. A feed’s page shows only items from that feed.

• Scroll down to see more items.

4. Install Apricot on the home screen

Apricot runs in the browser but is designed to look and feel like a native app. Like an Android or iOS app, Apricot can be installed on your phone’s home screen.

Most browsers will show a pop-up when you first navigate to Apricot that asks if you want to install Apricot on your home screen. If you miss the pop-up, not to worry, there is a button in most browsers’ settings to install a web app. In Chrome on Android, for example, look for the Install app option in the settings menu.

That’s it!

That’s all there is to it. Please give it a spin and let us know what you think. Feature requests, questions, comments—we’re happy to hear it all.

Happy browsing!

Social media killed the RSS star

For the past 15 years, I have been a data scientist, focused on machine learning and AI. Data science is a field that moves very quickly and it’s important but hard to stay up to date with things like new software, courses, conferences, papers, job postings, etc.

Once upon a time, I used an RSS reader to try to keep tabs on things. When Google Reader died, I—and many other data scientists—came to rely on Twitter to stay in the loop professionally.

For a time, the social media revolution worked well. The mobile-first, scroll-based UI was much simpler than dashboard-style RSS readers. People posted links to content on all sorts of platforms, not just blogs. The platforms’ ranking algorithms introduced us to things we wouldn’t have otherwise seen. Personal commentary added color, trustworthiness, and a sense of connectedness. It was puppies and rainbows, for a time.

Social media is bad

You know the sad end of the story; social media evolved into a toxic dumpster fire. Now, the good stuff is drowned out by clickbait, trolls, bots, and harassment. Driven by ads-based business models, social media platforms keep users addicted and track users across the web. Plus, it’s clear now that a handful of loud influencer voices does not truly represent a vibrant community.

Moving past social media

Apricot’s premise is that we can un-bundle social media, keep the good bits and throw out the toxic junk.

What’s the good stuff? Great question—glad you asked. Here’s what we should keep in the next evolution of media browsing:

• The user decides what sources make up their feed.

• Content from all sorts of web platforms.

• The user decides: smart ranking to see the most relevant items first, OR simple chronological order.

• Links that open in the original content platforms.

• A mobile-first design philosophy.

On the flip side, let’s get rid of:

• The ads-based business model that drives user addiction, mental health problems, and privacy violations.

• Harassment from bots, trolls, and other random internet strangers. I never want to have that sickening feeling while watching an army of trolls on the attack.

Apricot in a nutshell

Apricot lets you subscribe to content from various platforms on the web: YouTube, Substack, Reddit, Spotify (podcasts), and any RSS feed. Add your favorite sources and Apricot will import all the items from those sources as they’re published, sorting them so the best stuff is at the top. Like social media, you simply scroll to see more items. No enterprise dashboards here, browsing shouldn’t feel like work. Click items to read/list/watch on the original platform. Bookmark items to save for later. That’s it—super simple!

Where we’re heading

Today’s Apricot checks most of the boxes, but there’s a lot of room to grow. More content sources, better discovery of “unknown unknowns”, contextual item ranking, summarization of long-form content with state-of-the-art language models—everything’s on the table, as long as it stays true to our core principle of keeping control in the hands of the user.

Sign up at https://app.theapricot.io and check it out. Please don’t hesitate to send us feature requests and feedback—it’s early days, so your input will make a big difference!