Private markets are slow going, you’re not inundated with an onslaught of disclosures, earnings season is not something you lose sleep over, and stock price movements do not concern you. That’s especially true when you’re in tech and your days revolve around asking founders how you can be helpful. But whether you pull in all nighters tweaking that financial model or attending yet another event that ends way past your bedtime, you always need to read the news, even if you nod your head in approval when you bump into Joe Schmoe from All In Capital and they blabber about this hot new startup working on an agentic dev tool … and you have no idea what they’re talking about.
Public markets have it easy when it comes to news – at least on the micro level. Companies use standard terminology, they explain the jargon they throw in their disclosures, and the market chews them up if they stray from these “terminology standards” (see Wework’s community-adjusted EBITDA). But in a largely private tech ecosystem with enablers foaming at the mouth at the mention of AI, there are terms sprouting out left and right that their very definition is being disputed. "You'll see every other startup cram the word 'Agentic' into their logline (e.g., 'Agentic AI for hiring,' 'Agentic CRM for blue-collar workers,' or 'AI Support Agent orchestration platform' are some common descriptions)." Most of the time the buzzwords are noise, a distraction from the core business offering. That’s not to say public companies don’t attempt to bait their investors with this nonsense, but they have earning calls.
Checking the news is part of your daily routine – you get alerts on the latest fundraising rounds to your inbox, check the social feeds for the latest company announcements and maybe that launch video that went viral, and you listen to the mandatory podcasts (well you watch the clips). In a sense, you feel informed about all the latest tech happenings, but the big picture is unclear. And to get more context, you’re usually 5-10 google searches away and 30 open tabs from understanding what a single piece of news means, what the implications are, and what the backstory is.
Step 1: Getting the news
All the essential research and data platforms have their own newsfeeds that you can customize; they aggregate news from different sources and even report their own. But to build your own custom newsfeed, you’ll need to go directly to the source. We’ll go with the free, tried and tested RSS feed. Though not all news outlets maintain RSS feeds and the full content is not made available, it'll do for our use case. You can always integrate with a news aggregator if you’re willing to shill a few bucks.
Step 2: Automating the search for context
Running parallel web searches to understand the precedent of Cowabunga ai raising a significantly large seed round is an arduous task. You’ll need to look up other companies in the space but first define what even that “space” is, which is challenging given tech’s obsession with coming up with new terminology every few months. You’ll probably want to then search temporally, looking back more than 5 or 10 years ago to match that pattern “large fundraising round in frothy space” with a previous tech cycle – information that is buried in old press releases and academic papers.
When reading about fundraising rounds in frothy market times, there’s room for misjudgement, FOMO, but also confusion. Cowabunga describes itself as a “cursor for customer service.” Sure, you’ve heard about cursor, the agentic coding tool, but that description is just noise, it doesn’t explain what the business does.
A large language model can deconstruct that description down to common principles, so you get keywords including “automation” and “enterprise customers.” We’ll then automate the search process, passing the news articles to a large language model and asking it to generate search queries for us; queries stripped from jargon and grounded in structural business patterns observed in the article. The search engine runs those queries and returns a set of news articles.
Step 3: Finding analogues
With our source article, search queries – and in turn our search results – all discussing common business patterns, it now becomes easier to run pattern matching. The goal is to answer the question “what are the common business patterns observed across our search results?” This is where we aggregate all the knowledge gathered from the search results and reorganize it in pattern buckets. Each bucket is an explanation of a common business pattern observed in tangentially similar companies or industries to Cowabunga.
Step 4: Reed your feed
At this point, you will have deconstructed a news article down to common business patterns, searched for precedents, and aggregated all that research into business patterns – full context on a startup story all in a matter of seconds. Now all there’s left to do is just organize that context into a digestible newsfeed you can share with your co-workers. You’re just one prompt away from creating your own template.
This might all seem too laborious, but it’s a one and done setup for a daily contextual newsfeed. I’ll take that over 30 Google tabs any day.
You can try it out yourself by running this notebook