H
Harvest
AI Summarized Content

How this app saved me from infinite scrolling

This video shows how Reysu used Claude Code + Obsidian to build simple automations that cut down email checking, message overload, YouTube doom-watching, news scrolling, and paid subscriptions. The core idea is: don't just ask AI for advice—use AI to do the work, then save the results as searchable markdown notes you own. He also explains why most "AI summaries" feel like slop, and how to get high-fidelity summaries using full transcripts and "sub-agents."


1. From "productivity black hole" to fewer apps and less scrolling

Reysu starts by saying he's been in a "productivity black hole" lately—building a few small apps that made a surprisingly big difference in daily life. The outcome is simple but powerful: less checking, less opening apps, less scrolling, and fewer subscriptions.

He credits most of it to Claude Code (correcting the subtitle's "Claw/Clock Code" typo), and frames it as a tool with super high ROI—meaning the time and money you invest into setting it up pays you back a lot.

He contrasts it with typical AI chatbots. Normally, you ask something, get text back, and you still have to do the work. With Claude Code, it can answer and take action—like organizing files, researching, or generating learning materials—often without you ever seeing code.

"With Claude Code, not only does it give you an answer back, but it can also do the work for you."

He also points out a misconception: the name makes it sound like it's only for programmers. But he emphasizes it can be used for everyday life tasks—organizing your computer, researching, even making Japanese flashcards.

"The name Claude Code is a bit of a misnomer… you might assume it's only for programmers… but actually it can do a lot of things… completely unrelated to coding."


2. Daily Brief: email + messages without constantly checking them

The first concrete improvement: he stopped opening email and messages all day. The method is a Daily Brief—a single note generated each morning that contains only what matters.

Claude Code connects to:

  • Email (Gmail)
  • Messages
  • Calendar (Google Calendar)
  • Tasks (Things 3)

Then it writes a clean, daily note inside Obsidian (a local, markdown-based notes app). His Daily Brief includes things like weather, calendar items, and emails/messages that actually need responses.

"I have it create a daily brief every morning of the most important email and messages that I may have missed."

A key detail: it also clears the inbox by archiving spam, cold outreach, and low-priority messages—based on rules he trained it with.

"It archived some vague cold outreach, some spam… I've taught it to distinguish between things I want to be archived and things that don't."

This leads to an important realization: a lot of messages don't require any reply, or at least not an urgent one. Seeing everything in one place reduces anxiety and impulsive checking.

"A lot of the messages and emails I get don't actually need a response or don't need an urgent response."

And because the Daily Brief "surfaces" only what's important, he doesn't need constant notifications—and he avoids opening distracting apps like Instagram just to check DMs (and then getting pulled into Reels).

"I don't have to open up Instagram and be distracted by Reels…"


3. Wasting less time on YouTube with real, high-fidelity summaries

Next, he tackles YouTube time-wasting. He acknowledges many people already use AI summaries, but complains most of them are short, generic, and miss the nuance.

"Most of the time… it just generates a very short, sloppy summary that doesn't really capture the essence of everything that's said."

Instead, he built an Obsidian workflow to generate detailed, structured summaries of long videos and podcasts. He shows an example: a long Tim Ferriss podcast summary that includes:

  • A TL;DR section ("too long; didn't read")
  • Topic timestamps (so you can jump to the relevant part if you choose to watch)
  • A list of best quotes
  • A list of people and concepts mentioned

He emphasizes the biggest advantage: it's not a "wall of text." It becomes a dynamic document.

"The beauty of doing this in Obsidian is that it's not just a wall of text… it's actually a dynamic document."

Linking people and concepts (so you never feel lost)

When the transcript mentions someone (like Alex Honnold), Claude Code automatically creates a separate page summarizing who that is. Later, if another video mentions the same person, it links back—so your knowledge base becomes interconnected.

"When it's summarizing, [it also] make[s] pages for the things that are mentioned."

He shares another example: a multi-hour podcast featuring David Sinclair. Because he's already heard Sinclair before, summaries help him quickly decide whether the episode contains new information or repeats the same ideas.

"This really helps me save some time in figuring out: is there actually new information… or is he recycling the same talking points?"

He also notes a learning preference: he personally absorbs better by reading, because he can go at his own pace and look things up.

"I find it much easier to absorb knowledge when I'm reading it… because I can go at my own pace."


4. Why typical AI summaries are "slop" (and how he fixes it)

He explains the core technical reason many AI video summaries are weak: context window limits. In plain terms, AI can only "hold" so much text at once. With a 2–3 hour video, a chatbot often can't process the full content in one go—so it outputs a vague, safe summary.

"AI chatbots are limited by their context window… so… it just gives you a very generic summary."

His fix is to:

  1. Download the full transcript
  2. Split it into parts
  3. Use multiple sub-agents (smaller AI workers) to summarize each part
  4. Combine into a detailed, "Wikipedia-like" article

"The result is you have a very specific, Wikipedia-like article that you can just read through…"

Even better: if a video has no transcript, his workflow can download the video, transcribe it, then summarize it anyway.

"If there's no transcript, it will download the video, manually transcribe it, and then it will do the summary."

He mentions he packaged this as an Obsidian "skill" that can summarize not just videos, but articles, books (like EPUB/PDF), and podcasts, with summary length proportional to the original content.

"It summarizes it proportional to the length of the content."

So a huge book doesn't become a 10-line summary—it becomes something you can actually read and learn from in 15–20 minutes.

"You actually get the sense that you read a book or reread a book."


5. Understanding hard technical material by "going one layer deep"

He expands the same idea into deeper learning: using Obsidian + Claude Code to understand difficult research papers that normally require lots of prerequisites.

Instead of hitting a wall, he can ask for a detailed explanation written at his level, and click into any unfamiliar term to learn it. The system builds a connected "one layer deeper" set of supporting notes.

"Every single thing here I could go in and read more about it and see how it connects to everything else."

This is a subtle but important point: the summary isn't generic—it's tailored to what he's ready to understand.

"I had to write it in a form of writing and level that I could understand."

And because it uses multiple sub-agents and structured notes, he can get genuine comprehension rather than skimming and forgetting.

"I can actually get a deeper understanding of the paper."


6. Stopping news scrolling while staying informed

The third lifestyle shift: he stopped scrolling news feeds, but still stays updated. He uses a similar "Daily Brief" concept for news—Claude Code finds the top 3–5 most important news items and saves them in a daily Obsidian document.

"Find the top three to five most important news that I should be aware of… and then put them into a document."

He can expand past days if he wants, but the big win is speed: checking these pages takes 10 to 30 seconds.

"It only takes me 10 to 30 seconds."

He also customizes it by topic: not just world news, but AI, local Japan news, macroeconomics, and US news.

He explains the emotional benefit: feeds like Twitter/YouTube mix important news with endless distractions, which creates FOMO and pulls you into scrolling. A static daily doc removes that loop.

"It's really helped me not have to have any sort of FOMO… because I just have it in my Obsidian as a static document every single day."

He adds that this idea generalizes beyond news: you can have Claude Code continuously research a specific subject and keep updating one page—like a living document.

"It's basically a living and breathing document."


7. Cutting subscriptions by building tiny tools yourself

The fourth benefit is financial: he's stopped paying for certain subscriptions because he can build simple replacements. He says this is still mind-blowing even for him as a computer science graduate, because you're no longer limited by your own time ("man-hours") or technical expertise.

"You're no longer constrained by man-hours or technical knowledge. You can literally build anything that you can think of."

He gives a simple example: he wanted a time-lapse recorder for his computer day. Existing apps cost $5–$10 with mediocre reviews, so he asked Claude Code to build one—and it worked on the first try.

"I just told Claude Code to build it and it built it in one attempt and it works perfectly."

He explains the workflow: you describe what you want, it drafts a plan with you, you approve it, and it builds.

"All you have to do is tell it what you want… once you approve… it will just go out and build it."

He highlights how recent this is:

"This… literally was not possible even just one year ago."

He adds a realistic caveat: simple apps with one purpose work best. Very complex apps can get messy and bug-prone.

"This works better for simple apps… [for] a very complex app… it can become too unwieldy."

More examples: Obsidian extension + website hosting cost

He built a tiny Obsidian extension that shows an unread indicator when a note updates—done in minutes.

"That literally just took a couple of minutes."

He also reduced his website costs: he had been paying $50–$60/month using Notion as a CMS backend, but Claude Code suggested rewriting the backend and hosting on Cloudflare. After a few hours of work, hosting dropped to $0 with the same features.

"Now my website literally costs $0 to host… I have [an] identical feature set."

Because these savings add up, he says he's now on Claude Code's highest paid plan—because it's still net-positive financially.

"I've literally saved thousands of dollars with Claude Code."


8. Supercharging Japanese (and other hobbies) with automation

The fifth benefit is personal growth: using Claude Code to accelerate learning—especially Japanese. He recently moved to Japan, and calls out one of the most annoying parts of language learning: making flashcards.

He uses Anki, which he calls the best app to study anything, but creating flashcards manually used to take a lot of time.

"One of the most time-consuming things… is making flashcards."

Now he automates it: Claude Code checks his to-do app, and when he adds a new Japanese word, it automatically creates an Anki card by:

  • taking the word
  • finding an example sentence
  • pulling a definition from chosen dictionaries
  • generating the flashcard

"It'll take the word, find the example sentence, find the definition… and it'll just make the flashcard."

So his daily workflow becomes simple: open Anki and study, without spending extra time on card creation.

"All I have to do is open up the flashcard app and I can study new words every single day."

He mentions doing similar things for other hobbies like guitar and music theory.

And he makes a broader recommendation: if you're a self-directed learner, learning Claude Code is one of the best investments you can make.

"If you're a self-directed learner, learning how to use Claude Code is one of the best investments you can do today."


9. Why Claude Code + Obsidian together is the "underrated" combo

In the final segment, he gives what feels like his most strategic advice: if you use Claude Code, set up an Obsidian vault that Claude manages autonomously.

He personally keeps two vaults:

  • His own vault: human-written notes
  • Claude's vault: notes written by AI

He keeps the AI vault read-only from his side—he doesn't manually edit it. It becomes the place where summaries, project docs, meeting notes, and "profiles" live.

"I have Claude Code manage his own Obsidian Vault."

The key benefit: you're not locked into one AI provider. Because Obsidian notes are just local markdown files, you can switch models later—like if a better model appears, or if models run locally on your computer in the future.

"The beauty… is you are not tied to using Claude Code."

He calls it "super underrated" to let an AI-managed vault store memories, people profiles, meeting notes, and summaries. It can search quickly because everything is markdown.

"You can also see its knowledge base grow over time."

Then he drops a big framing line: this setup is basically owning your AI memories, because the knowledge is stored as files you control.

"This is basically you owning your own AI memories… they're literally just files on your computer."


10. A hopeful take on AI (and a quick wrap-up)

He closes by addressing "AI doom" content—people saying AI will take jobs and produce endless low-quality content ("slop"). His view is more optimistic and practical: used correctly, AI becomes an amplifier for what you care about, while reducing time spent on what you don't.

"If you use AI in the right way, it can really be an addition to your life…"

"It helps you spend less time on the things you don't care about and spend your time more efficiently on the things that you do care about."

He emphasizes that these tools aren't going away, so it's important to learn to use them well.

"They're not going anywhere… [learn] how to use them in a beneficial way…"

Finally, he thanks viewers, mentions it's his first video filmed standing (standing desk), and signs off.

"Hopefully this video has been helpful… I'll see you in the next video."


Conclusion

Reysu's main message is that AI becomes truly life-changing when it moves from "chatting" to "doing": generating daily briefs, cleaning inboxes, summarizing long content properly, building small tools, and automating learning. The backbone of his approach is Obsidian as an owned, searchable knowledge base and Claude Code as the automation engine. The result: less infinite scrolling, fewer subscriptions, and more time for focused work and hobbies.

Summary completed: 5/2/2026, 9:03:54 AM

Need a summary like this?

Get instant summaries with Harvest

5-second summaries
AI-powered analysis
📱
All devices
Web, iOS, Chrome
🔍
Smart search
Rediscover anytime
Start Summarizing
Try Harvest