DEV Community: Evan Dickinson

Learning vs Building

Evan Dickinson — Mon, 18 Aug 2025 07:35:22 +0000

There's two paradigms in programming: the one who knows how the stuff works, and the one who knows how the stuff works.

In the era of the "ship it, brah" culture, it's easier than ever to just build it, and that's evidenced by the new wave of builders in the past year or so. You can build a frontend without knowing a clue about react and build a (meh) backend hooked to a live db without knowing what "a supabase" even is. And, with a bit more motivation and scouring through some documentation, you can build a real NextJS web app. But, what do you really learn? Maybe with some ai help you can build an app, and you've learned enough about the framework to understand how pieces go together, but, is it optimal? Or, if time constraints say otherwise to refactoring, do you know how to optimize it?

I'm realizing there's perhaps levels, or two different, yet complimentary fields to programming. There's the practical; the skill of knowing what libraries to implement and what cloud software can get a product out fast, and with enough room to scale. Then, the theoretical; knowing how a computer works to some extent beyond mashing keyboard buttons and bright colors changing on a screen, and are the ones developing the amazing things like Bun and Nodejs. The current tech culture, by all means, favors the former. However it's the people like Ryan Dahl that get jobs at Google ;)

That, of course, isn't to say you should know how the assembler works before you write your first hello world javascript app. In this current climate, undoubtedly speed is everything, so you should be in your favorite ide as often as possible building your next million dollar SaaS. But it is to say, that the idea of making the "right" Cursor prompt doesn't scale, and eventually you will need to know things like what async/await means, and what blocking vs non-blocking I/O is.

Redis-Powered Crypto EdTech Content Pipeline: Scraping, Processing & Curation

Evan Dickinson — Mon, 11 Aug 2025 06:54:36 +0000

This is a submission for the Redis AI Challenge: Beyond the Cache.

What We Built

With Nodelet, when we set out to create the "Duolingo for Crypto Finance," we knew that content curation would be our biggest challenge. Crypto education content scattered across the web is dense, technical, and frankly boring for most learners. Our solution? A Redis-powered content aggregation pipeline that transforms fragmented crypto resources into engaging, digestible lessons.
While we started with Redis as a high-performance data layer, our architecture is designed to leverage Redis's advanced AI and streaming capabilities as we scale.

The Architecture: From Simple Cache to AI-Powered Pipeline

Our current implementation uses Redis as a critical buffer between our scraping services and LLM processing, but the real innovation lies in how we've architected for Redis's next-generation features:

1. Intelligent Backpressure Management with Redis Streams Our scraping jobs run 3x faster than our Qwen3 LLM processing can handle. While we currently use Redis as a simple cache, we're migrating to Redis Streams for sophisticated flow control:

python
# Future implementation with Redis Streams
r.xadd("content_pipeline", {
    "source": "coinbase_learn", 
    "content": scraped_data,
    "priority": content_score,
    "timestamp": time.now()
})

2. Content Deduplication with Bloom Filters Crypto content is highly repetitive across platforms. Our next phase implements Redis Bloom filters to eliminate duplicates before expensive LLM processing:

python# Planned: Probabilistic deduplication
if not bloom_filter.exists(content_hash):
    bloom_filter.add(content_hash)
    # Process with LLM

3. Semantic Search with Vector Embeddings The real "beyond cache" magic happens when we implement Redis vector search for content similarity and recommendation:

python# Future: Vector-powered content recommendations
r.ft("content_idx").search(
    Query("@vector:[VECTOR_RANGE $radius $vec]")
    .sort_by("relevance")
    .return_fields("title", "difficulty", "vector_score")
)

How Redis Scales Our Vision

Current State: Redis handles async scraping bottlenecks and data persistence between services.
Future Ready:
Redis Streams for real-time content processing pipelines
Vector Search for semantic content matching and personalized learning paths
Bloom Filters for efficient duplicate detection at scale
Pub/Sub for real-time learning progress notifications

Demo

This is just the first draft - but here we demonstrate leveraging AI and Redis to develop our first draft lesson templates

Technical Impact

By using Redis as more than just a cache, we're building infrastructure that can:

Process thousands of crypto articles daily without duplication
Provide instant semantic search across educational content
Scale to millions of learners with real-time personalization
Handle complex data flows with built-in backpressure management

What's Next

It was so much fun building this! We started with only two days left in the competition, so some of it is still rough around the edges.
Redis's AI-focused features are transforming how we think about content pipelines. Our simple caching layer is evolving into an intelligent content brain that can understand, deduplicate, and recommend crypto education content at scale.

Team Submission

Evan: evan.dickinson.flinn@gmail.com
Elliot: sonneselliot@gmail.com

New Perspectives on Development Speed

Evan Dickinson — Thu, 24 Jul 2025 02:39:53 +0000

This is a submission for the World's Largest Hackathon Writing Challenge: After the Hack.

Woo boy. Insane number of pivots. Burning 100k tokens per request. Foot constantly on the gas. That was my Bolt hackathon experience and, I think all of us can relate a little.

The Truth: Solo Development has changed - iteration > all

I was not much of a developer before AI, but even with it, I always personally emphasized understanding the code-base, using the ai only to explain, and developing fundamentals before writing a line. Now, all that is mixed into one, horrid amalgamation of ai, code slop, and speed.

It's truly unprecedented, but anyone (and literally anyone) can put out code, push it to a github repo, and deploy it on a live url, and have a functional website in 15 minutes. The barrier to coding has indeed, hit damn near rock bottom. An already saturated tech market has just had the flood gates burst open, and everyone and their mother CAN become the next tech millionaire. So, in this world, what is the key that I've gathered from the hackathon? Speed. Pure speed above all. Speed in ideas, in pivoting, in drafting MVP's, and tearing those MVP's down and creating a new MVP. It's become 10x more important now to understand what to build, then how to build it. Anyone knows how to build it - just ask your favorite AI tool.

The SaaS tech-bro seas have become more saturated than ever, and it only grows. It's the one who has the most adderall, the most ADHD, and the most autistic drive that succeeds in the internet world.

We are entering a new world, where humans outsource code to an AI Agent, and code is written and trashed at unprecedent rates. This hackathon has then, has taught me my first, and biggest lesson in the tech-bro SaaS underworld: Speed, speed, speed, and can we go even faster?

Best,
Evan