It's like a databank for your AI... Call, Retrieve, Reason over it, Make the next step. A good chunk of the prep work / calling / cleaning of data is now done. +10,200 alive companies, in Briefing form your AI actually understands. What call will you make?

Passed 10,000 individual alive companies across the world last week. USA/UK/EU/Japan/South Korea/India It is a switch decision if I ever go into backdating companies. But main focus is on serving alive companies because, well, they're doing something right to still be alive. A time for backdating the dead companies will arrive. Ideally first is to get the worldwide coverage of any and all live companies. Especially the messy, broken access, hard to read sources. It'll be interesting to see how wide and how deep the data will be servable in my care. PreReason'd.

Muse called me "one of our first developers on the platform" I saw the message for developers to submit connectors back in Sept. And I right away sent my submission Imagine Muse handling your finance ideas, PreReasoned? Any day now...

The main bit I've seen with India Company Data is the inconsistent reporting of documents. Visually. The way a document is scanned by a copier is a problem for most parsers. Unusual sizes, coloring, uneven lining and table dimensions. Reading the papers myself, I catch numbers/dates/statements that surprise me to have been sent out for public view. The complaint may be a blocker for most people. 100% applies for any analyst using AI to "web search" for such data. The nature of messy data; open for the opportunity to build out a streamline any serious researcher / action-taker needs to depend on. For any data caller whose strength is in Action. That's the PreReason prep work for you.

Websearch is great at finding sources, but does nothing after... How to access? Can it access? Oh, website is blocking bot access. Found documents, hard to parse, now needing to pay for a parser. All that is time spent / tokens spent, compared to a source who took care of the messy data AND Has the channels built for the LLM age of info searching. Past raw data pulling, that's where APIs are best.

Bit by bit progress with serving India companies in the same care as USA companies. The main challenge with this nation is more extreme than for EU/UK. For India, individual companies have to be found and contacted, one by one, to establish a consistent stream of data. Now imagine doing that across 100's (even 1000's) that exist in India. I'm sure there is a faster way, if I had 6-figures lying around every month. Before that happens, serving India companies through PreReason will be going through the detailed care of info depth, correctness, and serving-speed which is given to every other nation. Watch me.

An interesting case with European company data was the unique letters I found, manually traced, and cleaned out for LLM reading; Ørsted, L’Oréal, Mærsk. Also in keeping the currency numbers correct and specific to the company as they file in their country. (Rather than converting everything into Dollars) All to preserve the "raw" data for the true starting point when making the briefings / trend perspectives, especially when inflation or currency gain/devalue is cared about... Why stress the LLM, burning tokens on that work, when it can all be handled upstream... by the data provider? Feeling confident in doing the real work, section by section, of Europe until the job is done. Then to move on to the next area in financial markets.

Imagine that. Taking bets on a part of a company's performance, rather than the typical route of "betting on Earnings day". Like hedging on the smaller bits, to protect the "whole". The need for individual data sources of a company's numbers is seen... only revenue, only capex, only growth in a specific business arm. To then deliver such connected info chains at the speed of an API call + format best for LLMs. Let's make it happen.

Many styles of marketing to go through, when you have all the imagination in the world. Solo Founder may require many hats like Solo Modeling (a few times can't hurt).

When taking a stance on the market, you best come in with all fields covered. The prep work is done. One call. Every line answered. > fundamentals, covered > sources, dated > your call, with best foot forward

Not a fun feeling, when I found out about a very similar-service competitor surfaced to me, seeing their history of raising 9 figures, talent, and high focus on enterprise. On top of that situation, another take of someone mentioning the entire handing of AI; > no need for data providers > potentially no more need for terminals > no more visuals > essentially, AI handles all Everything condenses to prompting a theory, and let the code do all the work + send orders + manage orders. Indeed, to what was possible with finance + AI last year compared to this exact month... It was from reading a few articles on how Asia was all open with LLMs + Financial Data WHILE USA firms shunned the idea of anything AI touching financial services that got me interested in making my mark in this space. Time will tell if the "someone"'s idea is true from how fast AI has entered into the arena... > for about all businesses > ways of searching the internet > in what work is now delegated to machines vs. people As a Solo Founder, the moments of questioning the value of what you are building are painful... Yet a part of the journey towards great businesses. Ironic, going through pain to solve pain points in making pain killers. To read up all available info, break them down to core concepts, group what is theory and what is proven to have worked, then truly, at the end of the day, the next set of moves are all risks; a bet on yourself. And in the past few days, a better bet was made. Now the work needs to hold up that bet.

So much tension boiling up this weekend with AI-worries being back in the news. This time, the debt of key players/companies in the markets. > Who would have guessed front loading debt without a healthy ratio of proven revenue can be “an issue” when people want to exit. Data wise, an interesting mistake I still see AI make when reading a company’s filings is dates. When a report drops, when a debt schedule is submitted, anything officially sent to the government for the record. The smallest thing... time zones. The SEC’s own data sometimes shows New York time while labeling it UTC, a difference of 4 to 5 hours depending on the season. So results that came in at 4:03 pm, after market close, may get “reasoned” by AI as 12:03 pm. > More of a mistake when backtesting, creating a data trail, anything where number and labels matter among the millions of data points. A small time mistake, but it seeds the entire analysis as a flop. Now you are stuck figuring out manually what went wrong. The small details matter, especially when this type of data is the closest to influencing actual money.

As I am building PreReason, I work through a lot of moving pieces: data sources, connections, briefings, self-checks. The more time that passes, the more confident I got in this one idea. In financial data, conviction should come from the data, and not from the order button. I also kept seeing a divide. Platforms like @tradingview and @Bloomberg lead at what's made to be seen: > charts > dashboards > screens and news lines >> built for a person's eyes and judgment. In the age of agents, the other side matters just as much: what's made to be operated, the mechanisms that take an idea to action. Much of what I used was built around the first. primerkit is my step toward the second, for the agentic age of financial data. It's an environment built specifically for working with LLMs: > data organized for the agent > functions built for action > and connections that are easy to set up It's now public and open source. I'm very excited to see what this new way of working with financial data becomes for all of us.

Outside the top three markets (USA, China, Japan) do the reports and numbers live by their own unique standards. Bit by bit, they will all be figured out in a form much greater than raw numbers... raw metrics... ideal for LLMs to read over. The EU and UK has been a nice challenge in sorting out their numbers into, ideally, the same form as a US company would report. With PreReason, your model can call one figure, like a company's operating income, or the whole briefing at once. Your AI, with assistance from PreReason's MCP, is guided on what is best to call and reason over. Looking at you AI-token-bill watchers. So much more to decode inside the realm of Financial Market Data. The Region-Badge box needs to be completed.

Ever moving closer to what data analysts do that makes real value... Talking to their clients and going outside :) Sidebar; making these 4-image panels like an app service does in the iOS store puts a much high quality vibe than one big image. Need to experiment a bit more...

The UK and EU markets something. Filing systems + IDs no sane person remembers on the daily. Over PreReason's MCP, you can ask for a company like Rolls-Royce by name. No fancy codes. The power of dated/sourced data makes this all possible (and simple for everyone).

This feature took a good while for PreReason, and myself. But I'm happy it is live, running, and ever expanding. Curious to see what new findings can be reasoned over in the land of Financial LLMs.