> ## Documentation Index > Fetch the complete documentation index at: https://complete-docs-x17s80.allmind.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. # Thesis Monitor — Complete Guide > Register an investment thesis and let AllMind re-check it on your schedule — each review returns a verdict, what changed, the evidence on both sides, a recommended decision, and new ideas, with an alert the moment your thesis starts to break. Thesis Monitor puts a tireless analyst on every view you hold. You write down a thesis once — a trade, a company call, an event or macro bet, a theme — along with what would prove it wrong, and AllMind re-checks it on a schedule you set. Every review comes back with a clear verdict (is the thesis still holding, or is it breaking?), what actually changed since last time, the evidence on both sides, a recommended decision, things to watch, and fresh ideas. When a thesis weakens past the threshold you chose, you're alerted — so you hear about a broken thesis from AllMind before you hear it from the market. *Find it under **Research Workspace → Thesis Monitor**.* ## What it is A **monitor** is a saved investment thesis that AllMind keeps under continuous review. Each monitor has three parts: * **The thesis** — the original reason a trade, company, event, or theme is worth holding, and, critically, **what would prove it wrong**. * **A schedule** — how often AllMind re-checks the thesis (as often as hourly, or as seldom as once a month). * **An alert threshold** — how much the thesis has to weaken before you're notified. On each scheduled check, AllMind researches what's changed against new filings, earnings calls, and news, then returns a structured **review**: a verdict, the key developments, the confirming and disconfirming evidence, a recommended decision, watch items, and idea generation. Every review is kept, so each monitor builds a dated history of how your thesis has held up over time. Think of it as a **research diary that checks itself**: you record the thesis and the disconfirming signals once, and AllMind does the ongoing work of pressure-testing it and flagging the moment your reasoning starts to break. ## When to use it Reach for Thesis Monitor when: * You **hold a view and want it continuously pressure-tested** — a long or short position, a company call, a macro or event bet, or a sector/thematic idea. * You care most about **being told when you're wrong.** The feature is built around disconfirming evidence: it tracks the signals that would break your thesis and alerts you when they show up. * You want a **standing verdict and history** for a thesis — a running record of holding → weakening → challenged → broken, with the evidence behind each change — rather than a one-off answer. * You're juggling **many theses at once** and want a single place that surfaces the ones that need attention today. Use something else when: * You need a **one-time answer right now** → run it in **Agent Studio**, **Chat**, or **Reports**. Thesis Monitor is for *ongoing* tracking, not a single deep dive. * You want to **run a report, refresh a grid, or send a scheduled prompt** on a timer or around earnings → use **Automations**. Automations *produce and deliver* recurring output; Thesis Monitor *judges whether a specific thesis still holds* and maintains its verdict history. * You just want to **read the latest news or track a calendar** → use **News** or **Live Investor Relations**. Thesis Monitor and Automations look similar (both run on a schedule) but do different jobs. An **Automation** runs a task — a report, a grid, or a chat prompt — and emails you the result. A **thesis monitor** re-evaluates a *specific thesis* and returns a verdict on whether it still holds, tracking that verdict over time and alerting you when it degrades. Use a monitor when the question is *"is my thesis still right?"* ## How to use it ### Create a monitor Registering a thesis is a short, single-screen process — most fields are optional, and sensible defaults are already filled in. Open Thesis Monitor and choose **New monitor**. If this is your first one, you can start from a ready-made outline (Trade or position, Company thesis, or Theme) to get a proven structure. Choose **Trade**, **Company**, **Event**, **Theme**, or **Other**. The type tailors the starting outline and the default things AllMind watches for. Add a short title you'll recognize (e.g. *"NVDA margin expansion"* or *"Oil shock — energy long"*), then write the thesis itself. State the reasoning **and what would prove it wrong** — the disconfirming signals are what AllMind watches hardest. You can click **Insert outline** to drop in a structured template for the type you picked. Attach the **tickers** and **companies** the thesis concerns, describe the **trade or position** (e.g. *"long equity," "short puts," "watch only"*), and name an **event** if it's event-driven (e.g. *"CPI," "Iran war," "earnings"*). This focuses each review on the right names and catalysts. Choose your **alert threshold** (how weak the thesis must get before you're notified), the **review cadence** (how often it re-checks), any **email recipients**, optional **tags**, and **monitor notes** that steer what AllMind pays attention to. Leave **Monitoring on** so reviews run automatically. (See *Capabilities & key choices* below for what each option does.) Save the monitor. You can **Run the first review now** for an immediate baseline verdict, or wait for the first scheduled check. ### Read a review Open any monitor to see its latest review, laid out as skimmable sections you can jump between: * **Verdict** — the headline call (Holding, Weakening, Challenged, Broken, and so on), the quality of the evidence behind it, and a one-line summary. * **Decision** — a recommended action with its urgency, an optional due-by, the reasoning, and alternatives — so a review ends in *"what should I do about it,"* not just an assessment. * **What changed** — the key new developments since the last review, each tagged as supporting or pressuring the thesis, with sources. * **Pillars** — the thesis broken into its load-bearing claims, each marked confirmed, stable, weakening, at-risk, or refuted, so you can see *exactly which part* is moving. * **Evidence** — confirming and disconfirming evidence side by side, with links. * **Watch** — upcoming catalysts, data, filings, or price levels to keep an eye on. * **Ideas** — related trade or research ideas the review surfaced, with rationale, tickers, and the risk on each. * **Memo** — the full written analysis, which you can **export** as a document. ### Manage a monitor and its history From a monitor you can: * **Run review** — trigger a fresh review on demand, separately from the schedule (great for a baseline right after you create it, or an ad-hoc re-check when news breaks). * **Review history** — open any past review by date and compare how the verdict has evolved. Failed attempts are marked as such and never mixed into your analysis. * **Edit** any part of the thesis, coverage, schedule, or alerting. * **Pause / Resume** monitoring — a paused monitor keeps all its history but runs no scheduled reviews until you turn it back on. * **Copy link** to jump straight back to a specific monitor. * **Delete** a monitor you no longer need. Your list of monitors shows the **verdict**, when it was last reviewed and when it's next due, and puts the ones that **need attention** — anything alerting, or a failed review — at the top. You can filter (All, Needs attention, Running, Paused) and search across titles, tickers, and types. ## Capabilities & key choices ### The five thesis types A specific trade or position — the setup, why now, your variant view, expected confirmation, and what would make you wrong. A single-name business thesis — drivers, catalysts, main risks, and the evidence you need next. An event or macro thesis (a policy decision, a data print, a geopolitical shock) — the impact path, beneficiaries, losers/hedges, and the signals that would change the view. A sector or thematic idea — the best expressions, adoption/demand signals, and risks to the theme. Reviews lean into related-ticker idea generation. Anything that doesn't fit the above — a free-form thesis you monitor the same way. Each type comes with a starting **outline** you can insert into the thesis field, and it shapes the default set of signals AllMind tracks (for a trade: price action, earnings revisions, positioning, news flow, valuation; for a company: filings, calls, demand, margins, competition, estimates; and so on). You're never locked in — edit the thesis and **monitor notes** freely. ### Alert threshold — how early you want to hear The threshold sets how much the thesis has to deteriorate before a review raises an alert: * **Early** — alert as soon as the thesis starts **weakening**. The most sensitive; best when you want the earliest possible warning. * **Standard** *(default)* — alert once the thesis is **challenged** (materially in question). A balanced setting for most theses. * **Strict** — alert only when the thesis is effectively **broken**. The quietest; best for low-conviction or long-horizon ideas where you don't want to react to noise. ### Review cadence — how often it re-checks Pick how frequently AllMind re-evaluates the thesis. There are one-tap presets — **6 hours**, **12 hours**, **1 day** (the default), **3 days**, and **7 days** — or set any custom interval from **1 hour up to 720 hours (30 days)**. Cadence is an **interval measured from the last review** (e.g. "every 12 hours"), not a fixed clock time — so there's no time-of-day or timezone to set. Match cadence to how fast the thesis can move. A fast-moving trade around a catalyst might warrant every 6–12 hours; a long-horizon company or theme thesis is usually well served by daily or weekly. More frequent isn't always better — it's more reviews to read. ### Coverage, tags, and monitor notes * **Tickers & companies** focus each review on the right names. * **Trade or position** and **Event** add context (what you actually hold, and the catalyst in play). * **Tags** help you organize and search a large list of monitors. * **Monitor notes** are free-form instructions that steer every review — the specific metrics, sources, catalysts, disconfirming signals, and idea-generation angles you care about. This is the single most powerful way to tune what AllMind looks at. ### Monitoring on / off A monitor set to **Monitoring on** runs reviews automatically on its cadence. Switch it to **paused** and no scheduled reviews run until you resume — useful for parking a thesis without losing its history. (You must resume a paused monitor before you can run a manual review.) ### Alerts and where they reach you An alert fires when a review's verdict **first crosses — or worsens past — the threshold you set** (not on every run, and not again for a verdict that's already beyond it). The monitor is flagged **needs attention** at the top of your list, and the alert is delivered through the platform's notifications: * **In-app** — alerts and attention-needed events land in your notification inbox, tagged *Thesis monitor*. * **Email** — add **email recipients** to a monitor and they're emailed when it alerts. * **Text (SMS) & desktop** — if you've enabled these in your notification settings, alerts reach you there too. Thesis alerts count as important, so they still come through even if your SMS is set to **"failures and alerts only."** (SMS is opt-in and needs a verified mobile number.) If several scheduled reviews fail in a row, the monitor is also flagged as needing attention, so a quietly-stuck monitor can't slip past you. ## What it can access & produce **What each review draws on.** Each review is a genuine piece of **deep research** — a thorough, multi-step investigation that deliberately hunts for evidence *against* your thesis (every review runs at least one counter-thesis check) and focuses on what's genuinely new since the last review. It works from the platform's **public market-research universe** and the **live web**: * **Regulatory filings** — U.S. (SEC) and Canadian (SEDAR) filings. * **Earnings materials** — call transcripts and presentation slides. * **Company communications** — press releases and investor-relations materials. * **Independent research** — third-party broker/analyst reports and ESG research. * **Market & fundamentals data** — company financials, estimates, and pricing for the names you attach. * **News and sector data**, plus **live web search** for anything breaking or outside the corpus. **Thesis Monitor works from public market data and the web — it does *not* read your private documents.** A thesis review does not open your Data Room, your uploaded files, or your connected Gmail, Outlook, Google Drive, OneDrive, SharePoint, or Snowflake. If you want a thesis pressure-tested against your own models, notes, or internal data, run that analysis in **Agent Studio** or **Chat**, where those private sources can be turned on. **What each review produces:** | Section | What you get | | ---------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Verdict** | The thesis status — Holding, Weakening, Challenged, or Broken (plus Untested before the first review, and Inconclusive when the evidence is too thin to call) — with a plain **evidence-quality** read of **strong**, **mixed**, or **thin**. Reviews give transparent qualitative judgments, not invented confidence percentages. | | **Decision** | A single recommended action — hold, add, reduce, hedge, exit, wait, or investigate — with its urgency, an optional due-by, the reasoning, and alternatives. It's a judgment prompt for you, never an automated trade. | | **What changed** | The key new developments since the last review, each marked supporting, pressuring, or neutral, with sources. | | **Pillars** | The thesis split into its core claims, each marked confirmed / stable / weakening / at-risk / refuted. The set of pillars carries forward and is updated review over review, so you can track each claim over time. | | **Evidence** | Confirming and disconfirming evidence, side by side, with links. | | **Watch** | Upcoming catalysts, data releases, filings, and price levels to watch. | | **Ideas** | Related ideas tied to your thesis — adjacent names, hedges, pairs, event-driven angles, or alternate expressions — each with its rationale and its risk. | | **Memo** | The full written analysis — exportable as a document. | Every review is timestamped and stored, so a monitor accumulates a **verdict history** you can page back through. The exportable memo and the surfaced ideas are what most naturally feed your next step — a deeper Agent Studio run, a report, or a note to the desk. ## Tips & best practices * **Lead with what would prove you wrong.** The disconfirming signals are what the monitor watches hardest — the more precisely you state them, the sharper the alerts. * **Use monitor notes to tune each review.** Name the exact metrics, catalysts, and sources you care about; it's the fastest way to make reviews feel tailored rather than generic. * **Match the threshold to your conviction.** High-conviction, fast-moving trades → **Early**; long-horizon or low-conviction ideas → **Strict**, so you're not pulled in by noise. * **Match cadence to how fast the thesis can move,** not to how much you want to check. Daily or weekly is right for most fundamental theses. * **Run the first review immediately** after creating a monitor to get a baseline verdict you can measure future reviews against. * **Attach the tickers and companies** — reviews are sharper when they know exactly which names the thesis rides on. * **Pause rather than delete** when a thesis is dormant but not dead; you keep the whole history. * **Read the Pillars section** to see *which specific claim* is moving, not just that the overall thesis weakened — it tells you where to dig. ## Limits & things to know * **Review cadence runs from 1 hour to 720 hours (30 days),** measured from the last review. More frequent than hourly isn't offered. * **It works from public market data and the web, not your private documents.** See *What it can access & produce* — for private-data analysis, use Agent Studio or Chat. * **Reviews run in the background and take a little time** — usually a few minutes, though a complex thesis can take longer. You don't need to keep the page open; the verdict and any alert will be waiting, and you're notified when a review completes. * **A failed review is not fatal.** If a review can't complete, your monitor is unaffected: the previous verdict stands and the next scheduled review retries automatically. Failed attempts are marked as failed and never fold into your analysis or verdict history. If several scheduled reviews fail in a row, the monitor is flagged as needing attention. * **Partial reviews are held honest.** If a review can't finish in full, AllMind holds the verdict at *inconclusive* and labels the read as partial rather than overstating a conclusion. * **A review stuck running for over two hours** is treated as stalled; it typically recovers on the next scheduled check, and you can retry it manually. * **Paused monitors never run** — no scheduled reviews, and you can't run a manual review until you resume. * **Manual reviews are on-demand re-checks,** separate from the schedule; running one doesn't change your cadence. * **Each monitor keeps its 100 most recent reviews** (always retaining the latest completed and the latest failed), so a long-running monitor holds a deep but bounded history. * **There's no set limit on how many monitors you can keep,** and the feature isn't gated to a particular plan tier. * **Room to write.** A thesis can run to about 20,000 characters and monitor notes to about 12,000, so there's plenty of space to be specific. You can attach up to 50 tickers and 100 companies, and up to 50 email recipients, per monitor. ## How it works with other features Thesis Monitor sits alongside the rest of the platform and both consumes and feeds it: * **Agent Studio, Chat & Deep Research** — a thesis review runs the same kind of deep research as the platform's interactive AI, but scoped to public sources. When a review flags something — or when you want the thesis tested against your **own** documents, which a monitor doesn't read — take it into **Agent Studio** or **Chat**, where you can turn on your Data Room, uploaded files, and connected data. Conversely, a conclusion you reached in Chat can become the thesis you register here. * **Automations** — complementary, not overlapping. Automations run reports, grids, and prompts on a schedule and deliver files; a thesis monitor judges whether a specific thesis still holds. Use both: an automation to produce your morning report, a monitor to tell you when the view behind it breaks. * **Reports & Grids** — export a review's memo, or use a monitor's surfaced ideas and pillars as the starting point for a formal **Report** or a comparison **Grid**. * **News & Live Investor Relations** — the developments a review surfaces come from the same market news and corporate-events universe those features track; a monitor is the *thesis-aware* lens over that same flow. * **Notifications, email & SMS** — reviews and alerts flow into the shared notification system, so a thesis alert can reach you in-app, by email, by text, or on the desktop, exactly like the rest of your research activity. ## Example workflows ### 1. Stand up a monitored short and get alerted when it's working — or wrong *Goal: track a short thesis and hear immediately if the reasoning breaks.* 1. In **Thesis Monitor**, create a **Trade** monitor. Insert the trade outline and fill in the setup, why now, and — most important — *what would make you wrong*. 2. Attach the **ticker** and set the **position** to *"short equity."* 3. Set the **alert threshold** to **Early** (you want the first sign of trouble) and the **cadence** to **12 hours** (a fast-moving trade). 4. **Run the first review** for a baseline verdict, then let it run. 5. When a review comes back **Weakening** or worse, you're alerted — open it, read **What changed** and **Pillars** to see which leg moved, and act on the **Decision**. ### 2. From a broken thesis to a fresh idea and a report *Goal: turn a monitor's alert into your next trade and a write-up for the desk.* 1. A **Company** monitor flags your thesis as **Challenged**; its **Ideas** section surfaces a related name that benefits from the same shift. 2. Take that idea into **Agent Studio** or **Chat** for a full deep-dive to validate it. 3. Collect the supporting material in a **Data Room** folder. 4. Generate a **Report** from a template to write it up — and register the new idea as its own **thesis monitor** so it's tracked from day one. ### 3. A monitored book across a whole watchlist *Goal: keep a standing verdict on every position you hold.* 1. Create one **monitor per position or thesis**, tagging them by strategy or sector. 2. Give each a cadence that fits its horizon (daily for trades, weekly for long-horizon holds). 3. Each morning, open Thesis Monitor and use the **Needs attention** filter — it surfaces only the theses that alerted or changed, so you read what moved instead of re-checking everything. 4. For anything alerting, work the **Decision** and **Watch** items; pause the monitors whose theses you've exited. ## Common questions An **Automation** runs a task — a report, a grid, or an AI chat prompt — on a schedule or around an event, and delivers the output. A **thesis monitor** re-evaluates a *specific thesis* and returns a verdict on whether it still holds, tracking that verdict over time and alerting you when it degrades. Use an automation to *produce recurring output*; use a monitor to *know when a view is breaking*. **Holding** — the thesis is intact. **Weakening** — early signs of pressure. **Challenged** — the thesis is materially in question. **Broken** — the core reasoning no longer holds. You may also see **Untested** (no completed review yet) and **Inconclusive** (a review couldn't reach a confident call). Your **alert threshold** decides which of these notifies you. Alongside the verdict, each review gives a qualitative **evidence-quality** read — strong, mixed, or thin — rather than a numeric score. A review runs deep research over **public** sources: U.S. (SEC) and Canadian (SEDAR) filings, earnings-call transcripts and slides, press releases and investor-relations materials, independent broker and ESG research, company fundamentals and pricing, sector data, news, and live web search. It does **not** read your private Data Room, uploaded files, or connected Gmail, Outlook, Google Drive, OneDrive, SharePoint, or Snowflake. To pressure-test a thesis against your own models or notes, run that in **Agent Studio** or **Chat**, where private sources can be enabled. As often as **every hour** and as seldom as **once every 30 days**, with one-tap presets at 6 hours, 12 hours, 1 day, 3 days, and 7 days. Set whatever fits how fast the thesis can move. No. Reviews run in the background — usually a few minutes — and the verdict, plus any alert, is waiting when you come back. You'll also get a notification when a review completes. Nothing bad. Your monitor is unaffected — the previous verdict stands, and the next scheduled review retries automatically. The failed attempt is marked as failed and kept out of your analysis and verdict history. You can also retry it manually. A review raises an alert when its verdict **first crosses — or worsens past — the threshold** you set (Early / Standard / Strict), not on every run. Alerts appear in your in-app notification inbox, are **emailed** to any recipients you add to the monitor, and — if you've enabled them in notification settings — reach you by **text (SMS)** or **desktop** too. Yes — open the monitor and choose **Run review** for an immediate re-check, separate from the schedule. It's the usual way to get a baseline right after creating a monitor. (Resume a paused monitor first.) Pillars are the load-bearing claims your thesis rests on. Each review marks each pillar as confirmed, stable, weakening, at-risk, or refuted — so instead of just *"the thesis weakened,"* you see *which specific claim* moved and where to focus. Yes — **pause** the monitor. It stops running scheduled reviews but keeps the full thesis and all its history, so you can resume any time. Deleting removes it for good. Yes — each review's full **memo** can be exported as a document, and the ideas and evidence it surfaces are natural inputs to a **Report**, a **Grid**, or a deeper **Agent Studio** run. ## Getting help For help framing a thesis, tuning what a monitor watches, or building a monitored book across your positions, reach AllMind support through the in-app support option or your account team.