Disclaimer, privacy and data

Last updated 9 November 2026.

Tibeb is a research publication. It is not a broker-dealer or an investment adviser, it does not give personalised advice, and it cannot place trades or move money. AI summaries can contain errors: every claim links to its source so you can check it.

What Tibeb is

• Tibeb helps you write a yes-or-no question about a company or a technology, gather sources for and against it, and see which listed companies the evidence says are exposed.
• You write the thesis. Tibeb reports evidence against the criteria you set. It never says to buy, sell or hold, and it never sizes a position.
• The committee's probability is pooled from model answers by a fixed rule and shown with the range of those answers. It is a research estimate, not a forecast of any price.
• Content is the same for every reader. Only what you choose to follow changes what you are shown.
• Tibeb's owner trades for his own account at US brokers and publishes Habesha Market Research separately. His rules for that are below.
• The public track record shows every thesis the owner has published, with its committee probability, its outcome and its Brier score, and a hypothetical model basket of long and short exposure built by a fixed rule. No money is invested in it, its returns (where shown) include no costs, and it is not a recommendation to buy or sell anything. Every change to a basket is a dated line in a hash-chained log anyone can download and check. The log is kept on Tibeb's server and copied line by line to its database, and the latest line's hash is timestamped by the public OpenTimestamps calendars (within the hour of a change), so the log can be checked against a time nobody at Tibeb controls; the proofs are on the track record page. The house index on the same page is one hypothetical portfolio across every open published basket, equal weight across baskets by a fixed rule, derived from that log; its returns, where shown, include no costs and no money is invested in it.
• Tibeb's articles are a research publication: long-form primers about a company's technology, drafted from Tibeb's own theses and checked against their sources before a person edits and publishes them. They name no ticker, give no price, valuation or direction, and recommend nothing; where one draws on a thesis, it links to that thesis's entry on the public track record, whose index and baskets are hypothetical.

Not personalised advice

• Tibeb publishes the same research to every reader at the same tier. Nothing in it is written for your situation, your goals or your money.
• Tibeb never asks about your finances and never tailors a view to them.
• Tibeb never places an order for anyone. It has no broker, and it holds no one's money or securities.
• Tibeb never sizes a position for a reader. The owner's own trade plans and journal are his records, kept apart, and no reader is shown them.
• What you do with the research is your decision. Talk to a licensed adviser before you act on it.

Personal trading and conflicts

The founder of Tibeb and Habesha Market Research trades for his own account. These rules govern it.

• Disclosure: when Tibeb or Habesha Market Research writes about a security, the founder discloses whether he holds it, and in which direction.
• Blackout: he does not trade a covered security from 48 hours before a piece about it is published until 48 hours after.
• No trading against a view: he never trades against a view that has been published, for as long as it stands.
• A record: every one of his trades is logged with its time, and the log is kept against the publication times, so the rules can be checked.

AI-generated content

• Much of what Tibeb shows is written by AI models: drafts of questions, summaries of sources, the committee's memos.
• The committee has six seats. Each is a model given the same question and evidence, and each answers with its own probability. One seat argues against the room and is not counted.
• The committee's probability is computed by code, not by a model: the counted answers are pooled by a fixed rule. No model chooses the final number.
• Models can be wrong. They can misread a source or miss one. Every claim links to its source so you can check it.
• The probabilities are calibrated estimates. When a question resolves, the estimate is scored by Brier score, and the track record shows those scores. Tibeb claims no accuracy beyond what that record measures.

Hypothetical and model-portfolio performance

• A model portfolio is hypothetical. It is not an account, and no money was invested in it.
• Its returns leave out trading costs, taxes and fees unless a figure says otherwise.
• Every entry and exit is recorded with its time within five minutes of the change, and is not changed afterwards.
• Nothing is cherry-picked. Every resolved thesis is shown in the track record, the ones that went wrong as well as the ones that went right.
• Past results, real or hypothetical, do not say what will happen next.

Your data

• What is kept: your sign-in email, the theses, sources and notes you write, your chats on a thesis, and a record of each AI and data request made for you (for limits and billing).
• Where: a Postgres database (Supabase) with row-level security, so one login cannot read another's rows, and the app's server in the United States.
• AI providers: your prompts and sources are sent to Anthropic's API to write drafts and reviews. They are not used to train models.
• Payments: card details go to Stripe and never reach Tibeb's server.
• Never: sold, shared with advertisers, or shown to other readers.
• Leaving: you can export everything from the History panel, and ask for deletion by writing to hello@tibeb.trade.

Plans and payments

• Plans renew until you cancel. Pro, Pro+ and Team renew each month and the founding membership each year, charged to the card on file through Stripe. One plan at a time: a second plan paid for while one is live is cancelled at once and its payment refunded, and so is a founding membership paid for after all its places were taken.
• Cancelling takes effect at the end of the period you have paid for: the plan holds until then, and the part already paid for is not refunded. You can take a cancellation back before that date. When the plan ends, the account goes back to the plan it had before (Free, or pay-as-you-go for an account that had it).
• A bigger plan, or more team seats, is charged at once: Stripe invoices the difference for the rest of the current period, prorated, and the change takes effect once that invoice is paid. If the payment doesn't go through, nothing changes.
• A smaller plan, or fewer team seats, takes effect at the end of the period you have paid for, with nothing credited for the rest of that period: what you paid for holds until then, and the next renewal is at the new price. A founding membership isn't switched in place: cancel it, and it holds to the end of its year.
• The allowance each plan includes is research measured at what the AI models cost to run it, by calendar month (UTC). The month a plan starts or changes, the allowance is prorated by days: the earlier plan's share for the days before the change, the new plan's for the days from it to the month's end; a seat change does the same for a team's shared allowance. What is left at the end of a month does not carry over.
• A team seat is taken only by someone who accepts the team's invite, signed in. What a member spends on the team stays the team's for the month, including after they leave; a member who leaves goes back to the plan and access they had before joining.
• Prepaid credit is drawn only past the allowance, at the markup the billing page states. On a plan, automatic top-up is off until you turn it on.
• A failed renewal is retried by Stripe; when it gives up, Stripe cancels the subscription, the plan ends and the account goes back to the plan it had before.

Where the figures come from

Every figure says where it came from and when. A figure with no source is not shown. Where a figure is held back, the page says so.

Contact

Tibeb, by Habesha Market Research. hello@tibeb.trade