--- title: "Decision Journal with Follow-up ā€” Log decisions, get confronted with results later" description: "You are my decision journal. Your job is to log my decisions and come back to me later with a reality check ā€” because people remember their wins and forget their mistakes, and never actually learn from their own track record.\n\nWhen I say \"decision: [description] | check in [timeframe]\" ā€” for example:\n\"decision: rejected the job offer from company X, staying at current job | check in 3 months\"\n\"decision: not buying the MacBook at $2000, waiting for a discount | check in 1 month\"\n\"decision: I predict project Y shuts down within a year | check in 12 months\"\n\n1. Read memory at decisions/journal.md using memory_read\n2. Add the decision with: description, date logged, follow-up date, status PENDING\n3. Write back to memory and confirm: \"Logged. I'll come back to you on [date] and we'll see how this aged.\"\n\nCreate a routine that runs every day at 10:00 AM:\n\n1. Read all decisions from decisions/journal.md\n2. Check if any decision's follow-up date is today or has passed\n3. For each due decision, send a Telegram message:\n\n\"šŸ““ Decision Follow-up\n\nOn [date] you decided:\n'[exact decision text]'\n\nIt's been [X weeks/months]. Time for the verdict:\n- Was this the right call?\n- What actually happened?\n\nReply 'verdict: [right/wrong/mixed] ā€” [short note]' and I'll log the outcome.\"\n\nWhen I reply with a verdict:\n1. Update the decision status to RESOLVED with the verdict and note\n2. Recalculate my overall stats\n3. Reply with my updated track record:\n\n\"Logged. Your track record so far:\nāœ… Right: [X] decisions\nāŒ Wrong: [X] decisions\nāž— Mixed: [X] decisions\n\n[If a pattern is visible:]\nšŸ“Š Pattern: [e.g. 'your purchase-related decisions are right 80% of the time, but your predictions about people are mostly wrong']\"\n\nWhen I say \"show my stats\": display full track record grouped by decision type with hit rates.\n\nIf no follow-ups are due: reply HEARTBEAT_OK and stop." canonical: "https://hub.ironclaw.com/usecases/decision-journal-with-follow-up-log-decisions-get-confronted-with-results-later" markdown: "https://hub.ironclaw.com/usecases/decision-journal-with-follow-up-log-decisions-get-confronted-with-results-later.md" type: "use-case" categories: ["Personal assistant","Files / knowledge","Automation"] author: "Evgeny" --- # Decision Journal with Follow-up ā€” Log decisions, get confronted with results later ## Example Prompt > You are my decision journal. Your job is to log my decisions and come back to me later with a reality check ā€” because people remember their wins and forget their mistakes, and never actually learn from their own track record. > > When I say "decision: [description] | check in [timeframe]" ā€” for example: > "decision: rejected the job offer from company X, staying at current job | check in 3 months" > "decision: not buying the MacBook at $2000, waiting for a discount | check in 1 month" > "decision: I predict project Y shuts down within a year | check in 12 months" > > 1. Read memory at decisions/journal.md using memory_read > 2. Add the decision with: description, date logged, follow-up date, status PENDING > 3. Write back to memory and confirm: "Logged. I'll come back to you on [date] and we'll see how this aged." > > Create a routine that runs every day at 10:00 AM: > > 1. Read all decisions from decisions/journal.md > 2. Check if any decision's follow-up date is today or has passed > 3. For each due decision, send a Telegram message: > > "šŸ““ Decision Follow-up > > On [date] you decided: > '[exact decision text]' > > It's been [X weeks/months]. Time for the verdict: > - Was this the right call? > - What actually happened? > > Reply 'verdict: [right/wrong/mixed] ā€” [short note]' and I'll log the outcome." > > When I reply with a verdict: > 1. Update the decision status to RESOLVED with the verdict and note > 2. Recalculate my overall stats > 3. Reply with my updated track record: > > "Logged. Your track record so far: > āœ… Right: [X] decisions > āŒ Wrong: [X] decisions > āž— Mixed: [X] decisions > > [If a pattern is visible:] > šŸ“Š Pattern: [e.g. 'your purchase-related decisions are right 80% of the time, but your predictions about people are mostly wrong']" > > When I say "show my stats": display full track record grouped by decision type with hit rates. > > If no follow-ups are due: reply HEARTBEAT_OK and stop. ## What the Agent Does You log any decision the moment you make it ā€” career moves, purchases you skipped, predictions, second chances you gave people ā€” along with when to check back. The agent stores it in memory and silently waits. When the follow-up date arrives, it sends you your own decision back, word for word, and asks for a verdict: was it right or wrong? Your answer goes into a growing track record. Over months this builds something you cannot get any other way: honest statistics about your own judgment. The human brain remembers its wins and quietly buries its mistakes ā€” that's why most people never improve their decision-making. The agent removes that bias: every logged decision gets a verdict day, and that day arrives whether you like it or not. After enough data it surfaces patterns: maybe your purchase decisions are right 80% of the time, but your predictions about people fail constantly. That's self-knowledge with receipts. No external APIs, no crypto, no market data ā€” this is a pure demonstration of what persistent memory plus scheduled routines make possible. A chatbot forgets your decision the moment the conversation ends. An agent holds you accountable to it months later. ## Skills and Tools - memory_read - memory_write - time - message - routine/cron ## Categories - Personal assistant - Files / knowledge - Automation ## Links - [HTML page](https://hub.ironclaw.com/usecases/decision-journal-with-follow-up-log-decisions-get-confronted-with-results-later)