A Tale of Two Prompts: Using ScrollWise for Deep Research

It’s hard to explain to people just how revolutionary ScrollWise AI is going to be for independent researchers, journalists and content creators.

It’s very easy to show you, though.

This is A Tale of Two Prompts, a show-and-tell on just how much better RAG technology can make your research and content creation process.

Prompt 1: Cultural Research

I like to write about culture research, specifically online cultures like the effective altruists, rationalists/post-rationalists and various political simclusters.

I know, I’m a bit weird.

So, let’s start with the easiest possible illustration of how gamechanging ScrollWise can be for research and content creation on culture. I’m putting (Vanilla) Anthropic at a disadvantage here, I admit.

First, let’s prompt good ole Opus 4.5.

Prompt: I’m writing a blog on the decay of internet cultures. Can you give me some insight on what I could write about?

Response: Great topic—there’s a lot of rich material here. A few angles worth exploring: The platform consolidation thesis. The early internet was characterized by thousands of distinct forums, blogs, and communities with their own norms, aesthetics, and in-jokes. The migration to a handful of dominant platforms (Reddit, Twitter/X, Discord, TikTok) has created a kind of monoculture effect where communities that once had strong identities become subsumed into homogenized feed experiences. You could trace specific communities—like early Something Awful, specific subreddits, or niche Tumblr subcultures—and document how their distinctiveness eroded as they scaled or migrated. Algorithmic curation vs. organic discovery. There’s a compelling argument that engagement-optimized feeds fundamentally cannot sustain genuine culture. Culture requires friction, gatekeeping, shared references that take time to learn—algorithms optimize for immediate dopamine hits. The “For You” page model treats content as interchangeable units rather than artifacts of a specific community. Cory Doctorow’s “enshittification” framework is worth engaging with directly—platforms attract users, then creators, then squeeze both to extract value for shareholders. This maps cleanly onto the lifecycle of internet communities. The death of digital permanence. Communities existed on platforms that simply vanished—GeoCities, Google+, countless forums. The internet increasingly has no memory, which makes sustained culture impossible. Context collapse is another rich vein—when your boss, your friends, and strangers all see the same posts, authentic expression becomes risky, and everything trends toward blandness. What angle resonates most with the piece you’re imagining?

This is, objectively, a good response. It mentions known figures in the area of study like Cory Doctorow, trends like platform consolidation and the problems with algorithmic social media.

The problem is, though, there’s not a ton of depth there. This is a pretty general answer to an admittedly general question. Let’s see what ScrollWise AI offers up for the same exact prompt.

Prompt: I’m writing a blog on the decay of internet cultures. Can you give me some insight on what I could write about?

Response: Based on the provided sources, here are several key themes and insights you could explore in your blog about the decay of internet cultures:

1. The Collapse of Tech Subcultures The sources describe how various online subcultures are reaching unsustainable end-states. TPOT (a tech-focused online community) is collapsing because members who spent time “larping as ‘cracked developers’” are being exposed for lack of substance beyond basic coding posts and aesthetic tech photos Toward a Humanist Accelerationism. This represents a broader pattern where performative online identities eventually face reality checks.

2. The AI Content Pollution Crisis A major theme is how ChatGPT and other LLM-generated content is flooding platforms, with some people building “fragile empires” in the top 10 on Substack using AI outputs You. Guy I Went To High School With. This creates a “vile feedback loop which hyper-optimizes for lukewarm content, in the process, destroying niche digital communities (which are generally the source of content experimentation).”

1. Algorithmic Rage-Baiting and Media Pollution The sources detail how algorithms incentivize harmful content through engagement metrics. One example shows how a tech account pivoted to transphobic content after discovering it generated 500,000 impressions, then continued doubling down for dopamine hits and financial payouts Information Pollution and How to Fight It. This creates cycles where “you post something that gets people riled up, the anger drives engagement, the engagement pumps the post in the algorithm.”

2. The Venture Capital Degradation VC firms are “funding slop, porn, weapons and cheap wrappers” because “they are absolutely horrible at finding anything of value to put money into.” Examples include funding “a coding IDE whose literal stated soul purpose is to let you gamble, swipe Tinder and watch brainrot while AI writes code for you” Toward a Humanist Accelerationism.

3. The Loss of Authentic Digital Spaces The sources argue that platforms like TikTok and YouTube have become so censored and algorithm-driven that authentic expression is nearly impossible. The dilemma is that avoiding these platforms means “You’ll be writing on a typewriter” - completely cut off from reach and influence Information Pollution and How to Fight It.

4. Potential Solutions and “Humanist Accelerationism” The sources propose several approaches to combat this decay:

Building ethical alternatives with better design (avoiding infinite scroll, notification spam) Creating intentional “cages” or closed communities where content cannot escape Developing browser extensions to remove dark patterns Supporting decentralized platforms and the indie web Treating social media as a public good rather than private commodity The overarching narrative is that multiple unsustainable systems in tech culture are collapsing simultaneously, creating an opportunity for “humanist accelerationism” - building new, more ethical alternatives alongside the crumbling old systems Toward a Humanist Accelerationism.

Woah… That’s way more depth, and much more specific to my area of focus. That difference is from only 4 different sources judged as relevant across all the content in my Topics and Resources. The same exact model, the same exact prompt, wildly different outcomes.

The ScrollWise AI Difference Maker

The difference maker is this: Anthropic, OpenAI and Google have access to an enormous amount of data. They use that data to train their models, and those models are obscenely powerful.

That said, the problem arises when you need to push past the level of “just asking a general question” and start really drilling down into niche research focuses, or areas like AI and tech culture where the “scene” and data and research around it is changing constantly.

RAG, or retrieval-augmented-generation, is a solution that adds context to LLM queries by retrieving relevant data or resources to the query. The classic and simple way to do this is to run a vector similarity search, which transforms your query into a series of vectors (just think numerical data) that it can then use to search across a large batch of data to find relevant information. When the search returns something useful, it puts that useful text into your query before sending it to the LLM.

The reason this works so well is the LLM now has relevant, up-to-date context about what it is you’re researching. You get a much more specific answer, something much closer to what you’re looking for, than you would just running a general query across Claude, Gemini or ChatGPT.

But wait… it’s getting better

This week, we’re shipping some awesome new features, including:

• Private notes: store notes about a resource to add valuable context to the documents and URLs you save
• PDF uploading and text extraction: upload your favorite research papers and code documentation to ScrollWise and we will extract the text and make it searchable for you.
• API access: want to script your usage of the ScrollWise API? We’re rolling out API keys soon so you can start using ScrollWise with AI agents, REST API’s and more.

ScrollWise is only going to get more powerful. Getting in now will let you watch the meteoric rise from the front row while you power up your research.

Try It Free

We’re opening up ScrollWise to early users. You can save your research and use our chat and search functionaliuty for free, but if you want the unlimited experience, the Pro Tier is only $9 per month.

Power up your research for the price of a coffee.

Sign up today and stop losing your research.