Write something
Pinned
May 17 •
What you unlock at each tier
You joined as a free member. Read this whenever you get to the point of wondering what the paid tiers add. There are three rungs. Each one leads with a tool that does something no ordinary assistant does, and comes with the Classroom material that teaches you to use it. Nothing is held back to make the tier above look better. Free — $0 • Every community feed, including a daily market update, written for people building a system rather than trading one • 26 free tools. Some open with no account at all, and the rest are free too, they just ask you to create a free account. • Onyx, an AI assistant you can ask questions, on free models. Still in development • The Classroom intro modules, with every other module visible so you can see what is behind the wall before you decide Free is a real tier, not a trailer. If you are still working out the framework, it is the right place to stay. Premium — $19/mo in the community ($29/mo direct on the website) Premium unlocks Tephra, a private memory for what you know. • Tephra, a private memory for what you know • FinSim, a simulated paper-trading competition you can follow. No real trades are ever placed, and it is not advice • Script Lab and the Content Calendar • Member-only Classroom courses This is the tier for anyone actively building their stack out. VIP — $39/mo in the community ($59/mo direct on the website) VIP adds Onyx. • Onyx, an AI assistant you can ask questions. Still in development. It is an educational assistant, not an advisor, and it does not tell you what to buy • Every member-only Classroom course The website — same ladder, more ways to pay The tiers here are the same products the website sells. The community is the cheaper monthly door, which is the point of it. • Premium — $19/mo here, or $29/mo (or $290/yr) on the website • VIP — $39/mo here, or $59/mo (or $590/yr) on the website The community monthly is the lowest running rate there is. Annual is the website's discount against its own monthly price, not against the community rate. Annual, and buying Tephra or Onyx outright with a single payment, exist on the website only.
3
0
Pinned
Jul 21 •
A memory for the money stuff you keep re-learning
Most of us learn the same thing about our money three times. You read up on a rate, you work out your own situation, you make a call, and six months later you are half-starting over because none of it was written down anywhere useful. Tephra is the fix we have been building for that. It is a memory layer: you teach it what you know, your notes, the framework, the details of your setup, and it holds them and hands the right piece back when you need it. Not advice, not predictions. Your own knowledge, remembered and searchable, so you build on it instead of rebuilding it. It is part of Premium at the member rate. If the idea of never re-deriving your own money notes sounds useful, reply and tell me the one thing you always forget and have to look up again. That is exactly what it is for. Available in the App Hub (found in the classroom tools) Educational only. Not financial advice.
Pinned
Jul 21 •
Onyx: where it stands today
Onyx is an AI assistant inside the App Hub. Today you can ask it questions and get answers. It is part of VIP at the member rate, and free and Premium members can use it on free models. It is still in development; we will post here as each new ability goes live. Educational only. Not financial advice.
1
0
46m •
AI and Finance: how to spot a citation that was never real
A fabricated number is bad. A fabricated source is worse, because it does the one thing that is supposed to protect you. When an answer arrives with a named study, a regulation, a section number, or a publication, the normal reaction is to relax, and that reaction is the vulnerability. Sources are how we outsource verification, so a false one buys more trust than the claim could have earned on its own. These fabrications are convincing because they are assembled from real parts. The publication exists. The agency exists. The document naming convention is correct. The author is a real person who works in roughly that field. The year is plausible. What does not exist is that specific paper, that specific section, or that specific sentence inside a document that is otherwise real. The last case is the hardest to catch, because everything you can check at a glance is genuine and only the load-bearing detail is not. There are patterns. A source with no link at all is a warning, and a link that goes to a homepage rather than a document is the same warning wearing a coat. Section and paragraph numbers that are suspiciously tidy are worth a look. So is a citation that supports the sentence too neatly, stating exactly the claim being made with none of the qualifications real sources carry. And if you ask for the source twice in separate conversations and get two different ones for the same claim, at least one of them was produced rather than recalled. The only reliable check is the boring one. Open it. Not the search result, the actual document, and then find the sentence the answer attributed to it. If a regulation is cited, read the section. If a figure is attributed to an agency, find it on the agency's own page. This takes a few minutes and it is the entire value of having a citation. An unopened source is not a source, it is a decoration that made you stop asking questions. Worth adding that a tool being wrong here is not dishonesty, it is the mechanism working as designed on a request it cannot satisfy. It was asked for a citation, so it produced text shaped like one.
0
0
2h •
This week: ask a tool five financial questions you already know the answer to
Everyone using these tools for money is working from an impression of how reliable they are, and almost nobody has tested it. An impression formed from questions you could not check is not worth much, because the only cases you remember are the ones where the answer sounded good. This exercise replaces the impression with a small amount of evidence. Pick five questions where you already have the answer in front of you and can verify immediately. Good candidates are things you can read off a document. What a specific clause in one of your agreements means. What the total of a particular category was in a month you have already reconciled. What a term in a statement you have refers to. What charges apply to an account whose fee schedule you are holding. The requirement is that the answer exists in a document you can open, not in your memory, because memory grades generously. Ask each one in a fresh conversation, write down the answer, then check it. Score each as correct, correct but incomplete, or wrong. The middle category is the one to pay attention to, because incomplete answers are the common failure and the one that does not announce itself. An answer that is true but leaves out the charge that would have applied to you reads exactly like a good answer. Five is enough to see a pattern. Most people find high accuracy on definitions and explanations, lower accuracy on completeness, and the worst performance on anything requiring the tool to notice what you did not ask about. That is a useful profile to carry, and it is yours rather than a general claim from someone else. Keep the scores somewhere. Tools change, and in a few months you can run the same five questions and see whether anything actually improved, which is more informative than any announcement about a new version. After you score the five, how many came back correct but incomplete, and was that number higher or lower than you would have guessed beforehand? Educational only · Not financial advice · Results not guaranteed.
0
0
1-30 of 695
powered by
skool.com/obsidian-metrics-4683
Obsidian Metrics is an AI/financial education community for people building their own systems before they need them.
Suggested communities
Powered by