Built AIApplePie: Track AI news sentiment in real-time (Exciting vs Concerning)
Hey IH! š
Just launched AIApplePie (aiapplepie.com) mid January of this year - an AI news aggregator that lets the community vote on whether each story is "Exciting" or "Concerning."
The problem: AI news is scattered across multiple websites and sources. Hard to get a real pulse on how people actually feel about developments.
What I built:
- Aggregates 47+ curated AI news sources in real-time
- Community votes: Exciting or Concerning
- Clean, fast interface
Tech stack: Vue.js, Node.js, PostgreSQL, Redis, Socket.io
Next steps:
- Grow actual traffic (currently testing Reddit, X, now here)
- Build sponsored content feature for monetization
- Figure out if this thing has legs or not
Would love feedback! What would make you actually come back daily to an AI news site?
Such a nice job! Well done BRO!
Iām really interested on thisš
This is a really cool idea for a news aggregator! I love the concept of community-driven sentiment (Exciting vs Concerning) for AI news.
I'm currently building a similar engine called Lumen AI from Congo šØš¬, focusing on theological accuracy with a custom RAG architecture. Iād love to know: how are you handling the real-time sentiment analysis? Are you using a specific model to categorize the news automatically or is it purely user-based?
Congrats on the build! š
Hey! Thanks - Lumen AI sounds interesting, curious about the theological accuracy angle.
To answer your question: 100% user-based voting, zero AI categorization.
I wanted real human sentiment, not what a model thinks people would feel. The āExciting vs Concerningā split is subjective - what excites a builder might concern a safety researcher. Keeps it transparent.
Curious about your RAG setup - what specific challenges are you tackling with theological accuracy?
Good luck with Lumen! š
Thatās a very fair point about human sentiment! Subjectivity is exactly what makes community-driven data valuable. š
Regarding the RAG setup for Lumen AI, our biggest challenge is 'Theological Hallucination'. In a religious context, misquoting a text or mixing up interpretations can be a major issue.
We use the RAG architecture to force the model to prioritize specific, verified theological sources (Bible versions, commentaries, etc.) before generating a response. Itās basically a 'grounding' mechanism to ensure that the AI doesn't just guess, but actually 'reads' the source first.
How are you planning to handle 'vote manipulation' as you grow? Thatās always a fun challenge with community-driven apps!
Interesting concept.
AI news is everywhere right now. I'm curious ā what would make someone come back to this daily instead of just checking X or Reddit?
Do you think the community voting is the main hook?
Great question - and honestly, itās the challenge Iām wrestling with right now.
The hook: Community voting shows you how people actually feel about AI developments, not just whatās trending. Are we excited or worried? That sentiment angle is unique.
The problem: Needs critical mass to be valuable.
Iām betting that once thereās real voting data, it becomes the pulse check for AI sentiment - something you canāt get from X or Redditās algorithm. But I need to get there first.
What do you think - is sentiment tracking enough of a hook, or am I missing something?
That sounds like a really cool idea. AI news moves so fast that itās hard to keep track of what people actually think about new developments. A tool that shows real-time sentiment like āExcitingā or āConcerningā could be really useful for getting a quick overview. When I write short AI posts like this, I sometimes use Clever AI Humanizer cleverhumanizer to make the text sound more natural and easier to read.
Really smart angle - the Exciting vs Concerning voting adds a layer most aggregators miss. Sentiment tracking on AI news is going to become increasingly valuable as the pace of development accelerates. What's your monetization plan? I could see a premium tier for deeper analytics or alerts on sentiment shifts being really compelling.
Thanks! Yeah, sentiment tracking feels like the differentiator once I have real data.
Monetization plan:
ā Immediate: Sponsored content placements (already have an inbound inquiry from a PR agency)
ā Medium-term: Newsletter sponsorships once I hit 1K+ subscribers
ā Long-term: Premium tier with sentiment analytics, alerts, API access - exactly what you mentioned
Right now focused on getting to critical mass with users so the voting data becomes actually valuable. The analytics/alerts angle is perfect for that.
Curious - would you pay for sentiment shift alerts if the data was good?
Really like the Exciting vs Concerning voting angle ā it turns passive news reading into community signal. Iāve been deep in the AI space building tools myself, and sentiment tracking is something I wish I had when deciding which models and APIs to invest my dev time in. Curious: do you plan to add filtering by topic (e.g., LLMs, robotics, regulation)? That would make it even stickier for builders who care about specific verticals.
Weāve actually got that built already! The categories (Product, Business, Policy, Research, ML, etc.) are live filters - click one and you only see that verticalās news.
So you can already drill into ājust LLM developmentsā or āregulation newsā and see the sentiment split for each.
What would make it even more useful for you - more granular subcategories, or maybe saved filter combos?
Hey Chris, congrats on the launch!
A couple of ideas - not sure if these are good ideas, but hopefully some food for thought...
1. Build an AI "Fear & Greed" Index... Instead of just article-level votes, what if you aggregated the data into a daily overall sentiment score ( "Today the AI space is 65% Concerning")? If you graph that over time, especially around major model drops, it becomes a super interesting data point. Tech blogs might like that sort of thing...
2. Target specific niches: There are a few distinct sub-communities that would use this for completely different reasons:
AI safety watchers: They'd love a quick dashboard tracking "concerning" developments and keeping a pulse on the danger level.
Investors & VCs: They are definitely monitoring tech sentiment for market signals.
Builders & Devs: Looking at what the community actually finds "exciting" is a great way to figure out what use cases to build next.
3. A weekly newsletter wrapping up the "Most Exciting" and "Most Concerning" stories based entirely on your voting data would be a great way to bring the content to them (and a perfect spot for those sponsored placements you mentioned).
Really cool project, definitely think it has legs!
Hey Panic_Not_Needed_Yet,
Thanks so much - really appreciate the thoughtful feedback. These are actually great ideas, let me dig into each:
1. AI Fear & Greed Index - I love this concept. Right now Iām in a cold-start problem (literally the only one voting to make the site not look dead š ), but once I hit critical mass with real voting data, this becomes really compelling. Graphing sentiment over time around major releases (GPT-5, Claude Opus, etc.) could definitely be a unique angle that gets tech blogs interested. Adding this to the roadmap for when I have actual data to work with.
2. Niche targeting - This is super helpful framing. Iāve been thinking āAI enthusiastsā as one blob, but youāre right - these are three distinct audiences with different use cases:
ā AI safety folks would want a āconcerningā filter/dashboard
ā VCs/investors tracking sentiment as a market signal is brilliant - hadnāt thought of that angle
ā Builders using āexcitingā votes to validate what to build next makes total sense
I should probably build separate landing pages or views tailored to each of these personas. The VCs angle especially feels like it could open partnership/sponsorship opportunities.
3. Weekly newsletter - Actually have the infrastructure built already (double opt-in, Resend delivery), just need subscribers to make it worth sending. The āMost Exciting vs Most Concerningā format is perfect and ties directly into the voting mechanism. Great potential home for sponsored content too.
The catch-22 Iām in: Need users to get voting data ā Need voting data to attract users. But youāre giving me ideas on how to break out of that - the niche targeting especially.
Really appreciate you taking the time to think through this. Going to explore the VC/investor angle and start planning those persona-specific views.
Thanks again! š
I'm interested.
Looks useful. AI news moves fast, so real-time sentiment tracking could be really valuable.
One thing Iām experimenting with in a project is letting users try the product before signing up.
Thatās actually a really good point - friction at signup kills conversions.
Right now you can browse all the news without an account, but you need to sign up to vote.
Many thanks! š
does it account for fake news? AI generated political content? does it get filtered?
Good question. Right now Iām curating from 50+ trusted sources (TechCrunch, MIT Tech Review, ArXiv, official AI lab blogs, etc.) - no random blogs or sketchy sites.
So filtering happens at the source level, not content level. If a source consistently publishes garbage, I wonāt add it.
No AI content detection - but since Iām pulling from established tech publications and research sources, the risk is low. Political spam doesnāt really show up because those sources donāt publish it.
That said, if this scales and I open up submissions or add more sources, content filtering would become critical. Not there yet.