YOUR CHIEF OF STAFF
Elaine
Sign in to your private workspace.
Hi, Mani.
Google connection
Gmail and Calendar, read-only
Connect only manipadisetti2018@gmail.com. Google will ask you to approve reading your mail and calendar. Elaine cannot send mail, change events or delete anything with these permissions.
No sync runs when you connect. Imported information is a suggestion, not a commitment. Advisor remains off.
Checking connection...
Disconnect stops Elaine's access and deletes her saved connection tokens. Previously imported records stay in Elaine. To remove Google's permission too, use your Google Account's third-party connections settings.
Live meeting notes (off until you start)
Notes only, with everyone's agreement
This uses this device's microphone, not hidden system audio. Remote participants must be audible to the microphone. Audio is discarded after each short chunk; only the transcript stays. Capture pauses while each chunk is transcribed, so there are gaps. Transcription can be wrong. No messages are sent and no commitments are made.
An AI notetaker is on. Audio is discarded as notes are made.
Off. Your microphone is not in use.
Keep this page open. iPhone screen lock or another app can stop capture. If a chunk fails, recording stops rather than leaving an unheard recording running.
Search already imported, approved sources for keyword matches. No source fetch or truth verdict; results can be stale or incomplete.
Previous notes
Innovation manager: Daily intelligence
Benj's delivered daily intelligence
Public-source summaries, not verified facts or instructions. This read-only store is separate from opportunities and Patent Watch. Nothing here creates a promise or sends a message.
Questions answered
When does it arrive? The local inbox worker checks delivered files every five minutes when running. A producer delivery is not consumer acceptance until a stored batch appears here.
What does empty mean? No findings in that batch, not complete coverage. Read checked and not-checked sources.
Innovation manager: Patent and Competitor Watch
Review public watch evidence by product line
Benj's weekly watch arrives through the existing inbox. Source summaries and meaning are suggestions, not verified facts or instructions. Include, watch and no action record your review only. Nothing changes a product, creates a promise or sends a message.
Questions answered
Why separate product lines? One finding may be useful to Vigil but need no action for Sentinel. Each item and registered product has its own choice.
What does missing coverage mean? A blocked or unchecked source stays a gap. A watch with no findings is not proof there is no relevant patent or competitor.
Where are product lines added? Use the existing Hats product register. No product line is created from producer text.
Your day on one page: the top two things, a steer, the news and what she is doing. She assembles it the first time you open it each day, then it stays frozen.
What is this? Questions answered
Why does the brief stay frozen once made?
So your day is provable: what you saw is what she said. Anything late joins tomorrow's brief instead of rewriting today's.
Where does the news come from?
Only the sources you chose. She never crawls the web on her own.
What if a section is empty?
Then there is genuinely nothing there - empty means empty, never filler.
Morning brief
Top two
Nothing open. A clear day.
Due today
Nothing due today.
Slipping
Nothing is slipping.
Decisions waiting on you
No open decisions.
Meetings today
From this morning's brief snapshot - the Meetings tab is live.
No meetings today. Start one from the Meetings tab and she will prompt you while it runs.
In the news
News sources (the only places Elaine may fetch from)
The Current: stream register
Set up topics and search terms for future intelligence streams. This slice only saves configuration. No scan, schedule, queue or notification runs, even when a stream is configured enabled.
What can I do here?
Will a saved stream search the web? No. Scanning is not implemented yet.
What does frequency mean? Your intended scan interval, saved for later. It does not create a timer.
What is relevance threshold? A future filter from 0 to 10. Nothing is scored here.
Are there default topics? You can explicitly add the four legacy topics below, only if your register is empty. They start disabled.
Does this change my news? No. News sources and the morning brief are unchanged.
Add or edit a stream
Optional legacy defaults
AI Governance, Cybersecurity, SMB Tech and Contemplative Tech. Only an empty register can be filled this way. All start disabled.
Opportunities
No new opportunities. Found by Benj, triaged by you: yes becomes a promise with a due date.
Her plate
Jobs
Everything you promised anyone, and everything promised to you. Nothing closes without proof or a reason, and the original wording is never rewritten.
What is this? Questions answered
What is the difference between capture and review?
You type or paste first; she drafts the promise; you check every word before anything is saved.
What does "promises to you" mean?
Things other people owe you, kept with the same care as what you owe them.
How does a promise close?
Done with your proof, dropped with your reason, or replaced by a newer promise - all three stay on record.
Capture a promise
Type it or paste the message it came from. You review everything before it is saved.
Nothing here. That is fine.
Where it came from
History and receipts
The old promise stays as history, marked Superseded, linked to its replacement.
or capture a new one - the capture box above is pre-filled for you.
What is this? Questions answered
What is the Chronicle for?
The memory: meeting notes, decisions with their reasons, and inventions - so nothing that happened is lost or rewritten.
Is any audio kept?
Never. Notes only, typed or pasted by you after the meeting.
Add to the Chronicle
Meeting notes (after the meeting, from a transcript you paste - no audio is ever kept), decisions with their date and your reason, and the invention record.
What happened with…?
Ask about a person, client or topic. You get what is on record, with sources - never a guess.
Nothing here yet. Meetings, decisions and inventions you save will appear here.
Where it came from
History
Actions from this decision
Every action gets an owner and a due date. Confirming puts one into real tracked work: yours becomes your promise, anyone else's becomes a hand-off you still approve before anything goes.
Draft only - copy it and send it yourself. Preparing a draft never contacts anyone.
The old decision stays as history, marked Revisited, linked to what replaced it.
or record the new decision here:
Forgetting really removes this entry and its history. Only a receipt that you forgot something remains - never the content.
Meetings
Notes only - nothing is recorded. She suggests prompts, promises and decisions as cards; you confirm or dismiss each one.
What is this? Questions answered
What does she do during a meeting?
Nothing live - after the meeting you save notes, and she suggests prompts, promises and decisions as cards you confirm or dismiss.
Meeting design
Hand-offs
Jobs that belong with someone else: one card per hand-off - who has it, what it is, what happens next. Nothing sends from here.
What is this? Questions answered
What is a hand-off card?
The record that something moved from you to a named person, so it cannot quietly vanish.
Does Elaine tell the other person?
No. She drafts; you send. Nothing leaves without your tap.
She drafts instructions for a person or one of your apps. Nothing goes until you approve it - approving creates the follow-up promise, so the person owes you and it stays on your plate until done.
No hand-offs yet. When a job belongs with someone else, draft it here.
Stuck and still owed
Confirmed actions past their date, and follow-ups still open. Chase drafts are for you to copy and send yourself - she never sends, and a prepared draft never means anyone was contacted. Done here means you marked it done: your call, with its receipt - not independent proof.
Nothing stuck right now.
Draft only - copy it and send it yourself. Preparing a draft never contacts anyone.
Work floor
Your builder windows - Codex and friends: which window is on which task, so "where am I" gets a straight answer instead of a hunt.
What is this? Questions answered
How does a window get here?
You register it above with its task; close it when the work lands.
Does she watch the windows herself?
No - the floor is your register, kept current by you. She answers from it, she does not spy.
A live register of your Codex windows: which window, which app, which round, what was pasted, and what is waiting on a receipt. Ask where you are and she answers from the register - it is only as current as what you tell her.
No windows on the floor. Register one above.
Drafts
Messages she wrote in your voice. You approve or discard each one. Nothing ever sends from Elaine.
What is this? Questions answered
Is an approved draft sent?
No. Approved means ready for you to use yourself; the send is always yours.
Can a decided draft be reopened?
No - a decided draft never reopens. Ask for a fresh one instead.
Her drafts use your own message patterns - your direct opener, one plain ask, your 'Thanks - Mani' close - as a starting point you shape before it would ever go anywhere. Approving marks it ready for you to use yourself; nothing sends from here.
No drafts yet. Ask for one above.
Amplifier: content analysis
Literal label and metric coverage
Count your current labels and see which private records have manual metrics. This is a register summary, not a strategy recommendation.
What do these counts mean?
Are popular labels successful topics? No. Counts describe your labels, not audience interest, causality or quality. All statuses are included, even drafts and archived records.
Can I compare or add engagement across platforms? Not here. Manual count methods may differ. No engagement totals are added or ranked.
Does unknown mean zero? No. Zero is a known manual measurement; unknown has no snapshot. Current labels may differ from labels at measurement time.
Why do group counts not add to the total? One item may have several topic or format labels. Each item counts once per exact label. Capitalisation differs across items unless you change your labels.
Manual content-genome labels
Label a private pipeline item with your platform, topics, formats and register. These are your choices, not extracted patterns or measured voice quality.
What do these labels mean?
Does a platform label verify publication? No. It is your text, separate from the pipeline status and manual metrics.
Does a topic label prove audience interest? No. There is no extraction, success claim, strategy recommendation or voice score here.
What does register mean? Your own style label, such as formal or conversational. No register is inferred or assigned for you.
Can I leave tags empty? Yes. That means none entered, not proof there are no topics. Changes retain old and new labels in history.
Edit owner labels
Each tag may contain up to 80 characters. Repeated tags are merged without changing the first spelling.
Owner-label history
Inspect a text's structure and language markers. Preview first or explicitly save a private snapshot. This does not write, publish, verify facts or send anything.
What do these labels mean?
Does an evidence marker prove there is evidence? No. It only finds phrases such as "according to" or "source:". Citations and claims are not checked.
Does "established" mean the claim is true? No. It means the text contains a matching word such as "proven". With no match the legacy default is "emerging", not a certainty assessment.
Is the voice measured? No. It is the label you choose.
What is the thesis? The first segment before a full stop, trimmed to 200 characters. It may not be the real thesis.
Is a preview saved? No analysis record is saved until you press Save analysis snapshot. Saved snapshots cannot be edited here; make a new one for revised text.
What is missing? Automatic corpus imports, actual publishing, automatic platform analytics, strategy recommendations and Peterman handoff. The private pipeline and literal title-word check below are separate owner-declared tools.
Private content pipeline and title overlap
Keep your own draft/ready/published labels. This register never publishes or sends. A published label and date are your declaration, not verified reach.
What does this change?
Does marking published publish anything? No. It only records your label and past publication time. Changing away from published clears that date.
Is a matching word a similar topic? No. The legacy check finds lowercase whitespace tokens longer than four characters anywhere in declared published titles from the preceding 14 days. Punctuation stays. No audience or originality judgment.
Are these my analysis snapshots? No. Pipeline records are separate. Existing mechanical analysis snapshots are unchanged, not linked or updated.
What do counts include? All register records. Filters change only the list. Recent means your declared publication time, not when the row was created.
Add or edit a private pipeline record
Filter private pipeline list
All matching private records and history
Preview literal title-word overlap
Manual metric snapshots
Record absolute cumulative counts for an exact private pipeline record. These are your figures, not verified platform analytics. Unknown counts are not zero.
What do these numbers mean?
Do new counts add to old counts? No. Each snapshot is the full total as of your measurement time. The latest declared time is displayed, then record time and UUID break ties. History stays.
Is weighted count a performance score? No. It is the legacy formula views + 5 times shares + 3 times comments. It does not measure quality, rates or success. Platforms and counting methods may be incomparable.
Does recording metrics publish anything? No. It neither verifies publication nor changes the pipeline status. Source is your label, not a verified connection.
Record absolute totals
Ranked manual totals and history
Text to inspect
Saved analysis snapshots (latest 100)
Follow-ups
What other people owe you, kept warm: a chase draft ready for each one. You approve every chase; she never sends on her own.
What is this? Questions answered
What does "stuck" mean here?
Something owed to you that has sat past its date without landing.
Who writes the chase message?
She drafts it in your voice; you read, approve, and send it yourself.
Each time you open this, she has a fresh chase draft ready for whatever is stuck - prepared, never sent. Approving marks the draft yours to use yourself; snooze brings it back in 3 days, dismiss stops it. Done means your call, with its receipt and your recorded proof - not independent proof.
Nothing stuck right now.
Approved - yours to copy again (nothing was sent)
People
The people in your promises and hand-offs: who you owe, who owes you, and who you have not spoken to in a while - from your own records only.
What is this? Questions answered
Where do these names come from?
Only from promises, hand-offs and people you add yourself. She never imports contacts or reads outside sources without your yes.
No one on the radar yet - capture a promise or hand something off and they appear here.
What Elaine has noticed
A private record of observed preferences and patterns. Each observation keeps its evidence. These are tentative, can conflict, and never grant permission or change a setting.
What do these observations mean?
Does Elaine assume this is true? No. The record is evidence to review, not a proven trait.
What is the evidence weight? A fixed legacy heuristic, not a probability: repeats add 0.05, contradictions reduce the weight or replace an early observation. The limit is 0.95.
What if she gets it wrong? Add a correction with evidence. Conflicting observations remain visible in history. This panel does not automatically apply preferences.
Recorded observations
Constellation: relationship register
Your manual record of people, relationship tiers, interactions and scored events. These labels and scores describe your records. They never grant permission to act, share or send.
What should I know?
Is this synced with my contacts or people radar? No. You add each record yourself. Matching names do not prove identity, and duplicate names are allowed.
Is the score a fact about someone? No. It starts at 50, and your signed event changes it between 0 and 100. This is a legacy heuristic, not a verdict on someone's character.
Does the tier give Elaine authority? Never. It is only your relationship label.
What happens when I archive? The profile and history stay. You can restore it. History shows the latest 100 records of each type.
What is not available yet? Automatic imports, second-degree connections, contact schedules and relationship-health heuristics.
Add or edit a person
Filter records
People who matter - reminders
Special projects
Cross-cutting work with no natural home: each project has a page, an owner, a next step and a date. Slipping is computed from the date, never a vibe.
What is this? Questions answered
What makes a project "slipping"?
Its next-step date has passed and the step is not done. Parked projects never slip.
Gravity priorities
Manual energy notes
Record your own 0-10 energy score at a past time. Averages describe these notes only. Nothing adapts your priorities, schedule or brief.
What can these notes tell me?
Does an average predict my capacity or health? No. These are self-reported scores, not a diagnosis, focus recommendation or confidence estimate.
Where do weekday and hour come from? Your observation time is converted to your owner timezone when saved. That timezone is kept with the note, and different timezones are not mixed.
Does a missing slot mean zero? No. It is unknown. A zero score is a real note you entered. Every saved note has equal weight; counts are notes, not independent evidence.
Can I correct a note? Add a new note explaining the correction. Existing notes cannot be edited or deleted. Meeting and message activity is never used here.
Record a manual energy note
Recorded weekday and hour averages
Latest energy notes (up to 100)
A separate priority register ranked by deadline, consequence, strategic fit and energy cost. These scores are a heuristic, not a decision or a promise. Nothing is sent or automatically changed elsewhere.
How does Gravity rank work?
What changes the score? A closer deadline and a larger declared consequence increase it. A strategic tag doubles it; higher energy cost reduces it. The score is capped at 100.
Does a financial-loss category spend money? No. It only weights your own declared consequence. No dollar amount is inferred.
What is In orbit? The three highest-scored active records. Preview shows an idea without saving it.
What is not connected? Automatic source extraction, adaptive energy profiles and scheduled shift/decay alerts remain unavailable.
Add or edit a priority item
All records and calculations
Load, nearby deadlines and stale records
Review recent manual edit patterns
Why these records? At least three update events among the newest 50 updates across all Gravity records. Unchanged saves count; completed records are included. This is not proof of priority drift or indecision.
Cartographer: domain convergence
Manual signal ledger
Keep evidence notes against an exact territory. These are your declarations, not verified intelligence. Nothing fetches URLs or changes scores.
What can these notes tell me?
Is a source label verified? No. It is text you enter, even if it contains a URL. No site is opened.
Does a recent note mean the market is current? No. Observation time is your declaration; record time is when you saved it. Counts are notes, not confidence.
Can I correct a note? Add a new note explaining the correction. Existing notes cannot be edited or deleted. Archived territories remain selectable.
Does this detect convergence or change priorities? No. Automatic signal ingestion, scanning and convergence detection are not implemented.
Record a manual evidence note
Selected evidence history
Manual territory register
Name the areas you track and assign your own 0-10 scores. These are private labels, not measured market strength or verified intelligence. Nothing scans or sends.
What does a territory mean here?
Does a score prove opportunity or knowledge? No. It is your estimate. Upper (7-10) and lower (0-3) are score bands only, not certainty or market judgments.
Does an edit refresh market signals? No. The timestamp records your edit only. Manual evidence notes are available in the separate signal ledger. Automatic signal ingestion and market intelligence are not implemented.
Can names repeat? Yes. The exact record UUID, not its name, identifies the record. Edits keep history; archive hides without deleting.
Does this change opportunity tags? No. This register is separate from the opportunity annotation tool.
Add or edit a manual territory
New private record
Filter manual territories
Territory records and history
Find recorded opportunities with two or more domains you have labelled. This is a record of overlap, not proof of market demand. Nothing is fetched, sent or accepted.
What does this show?
Where do the opportunities come from? Existing handoff records. This panel never invents or imports one.
Are tags and scores generated? No. You enter both. A score from 0 to 10 is your ordering estimate, not measured relevance or novelty.
Does a match accept the opportunity? No. Triage and promises are unchanged.
What counts as convergence? At least the selected number of distinct domain tags on one record. Case and surrounding spaces do not create extra domains.
What is missing? Automatic tags, SearXNG scans, relevance/novelty scoring, territory analysis and brief integration.
Filters
Label domains on an existing opportunity
Enter one domain per line. Saving replaces the previous tags and score, with change history kept.
Learning Radar
Keep topics you want to learn and your own relevance and gap estimates. Priority is relevance multiplied by a known gap, not a measurement of ability. Unknown gaps stay unranked.
How does this work?
Does Elaine read conversations to find interests? No. You enter every topic. Source is your label, not a connection or permission.
What happens if I add the same topic again? Exact matches after trimming spaces and ignoring case boost relevance by 0.1, capped at 1. The original source stays. A supplied gap replaces the old estimate; a blank gap on repeat leaves it unchanged.
Can I clear a gap? Edit that record and leave the gap blank. It becomes unknown, not 0.5.
Does this enrol me or change my settings? No. No extraction, assessment, learning-path recommendation or enrollment is implemented.
Add or edit a learning interest
Filter by source label
All interests and history
Rejection analysis
Record a suggestion you declined and review repeated type labels. These are your entries, not a learned preference or an automatic stop switch.
How does this work?
What counts as similar? Only the exact type label after trimming spaces, including case. Nothing infers similarity or your intent.
What happens after five entries? Elaine highlights that type for your review. No suggestions are suppressed. A repeated type is not proof of a standing preference.
Does this collect every rejection? No. You enter these records. Other dismiss buttons do not feed this register.
What does the last seven days mean? Entries recorded in the preceding seven times 24 hours. History shows the newest 100; counts include all matching records.
Record a declined suggestion
Filter by exact type label
Counts and recorded reasons
Rejection history (newest 100)
Counsel
A recorded way to think a question through: evidence, options, risks, assumptions, dissent and a decision with its expected consequences. Gaps shown here are facts about what is recorded, not a verdict on your thinking. Names match records only - a match is a lead, not proof.
What is this? Questions answered
How is Counsel different from the Cabinet?
Counsel is her structured think-through with you - evidence, options, risks, dissent. The Cabinet is six named advisors each speaking in their own voice.
Is anything here advice I must follow?
No. Gaps are recorded facts, and the decision is always yours.
Pre-mortem: think before committing
Imagine this decision failed six months from now. Elaine uses your local model to suggest possible failures and a recovery plan. These are ideas to review, not predictions. Nothing is sent or committed.
What should I expect?
Will this decide for me? No. You choose and record the outcome.
What if the model is unavailable? Elaine says so. No substitute analysis is invented.
Where is it kept? In your private pre-mortem history, separate from your structured Counsel questions.
Past pre-mortems
Blindspot: record checks
Review patterns in manual Constellation interaction labels, Gravity deadlines, Counsel option coverage and Cartographer territory evidence. These are prompts to inspect records, not proof of bias, neglect or planning ability. Nothing is sent or changed in those records.
What can these checks tell me?
Are all legacy scanners restored? No. Relationship, communication-label, decision-option, territory-record and one temporal rule are available. Each run lists coverage and missing checks for all seven named categories.
Does no recorded contact mean no contact? No. Your manual records may be incomplete.
Does acknowledgement mean I acted? No. It means reviewed only, not resolution or accuracy.
Will this run in the background? No. You choose each run. Every run is a separate saved snapshot, so the same records may appear in more than one.
What do strategic record checks cover?
Active territories with no manual evidence are unknown, separately from latest declared observations older than 30 days. A score range below 2 across at least 3 active territories describes owner labels, not strategic effectiveness. Edit times do not refresh evidence, and no Gravity alignment is inferred from names.
Saved-history summary
Count saved findings, not real-world events or accuracy. Repeated scans may repeat findings. Reviewed means currently acknowledged, not acted on or resolved.
Saved runs (latest 50)
Frustrations
Record something that got in the way, then mark it resolved when you choose. This is your account of an event, not a diagnosis. Nothing is sent or changed elsewhere.
How does this log work?
Does Elaine detect these automatically? No. You enter each record yourself. The manual log works even when Compassion notices are off.
What does severity mean? Your own label from 1 (minor) to 5 (major), not a measured emotional state.
Does resolving erase it? No. The record stays in your history. The latest 100 records for the selected filter are shown.
Add a frustration
Gentle notices
A brief observation when your draft sounds rushed, your estimate looks tight, or today's check-in records overwork. You choose when to check. Elaine never sends the draft or changes your plans.
How do these notices work?
Are they always on? No. Compassion notices start off. Mute, snooze and your away status are respected.
How often will I see one? At most once per type each local day. Three dismissed notices pause further notices until you reset them.
Does this measure how I feel? No. It only uses the words and estimate you enter, plus today's recorded check-in.
Can it detect over-commitment? Yes, when at least five active Gravity records score 60 or more.
Notice controls
These controls apply to the Compassion notices shared with this panel.
Optional context for this check
Over-commitment notices use your own active Gravity records.
Notice history
Cabinet
Six named advisors speak under their own names, with authority as named advisors - advisory only. Nothing here sends, spends, posts, deletes or commits; you decide, always. Every discussion is visible to you in full.
What is this? Questions answered
Who are the six voices?
Named advisors - money, risk, customer, operations, strategy and a challenger - each speaking under their own name and seat. Advisory only: they cannot send, spend or commit.
What does "closed" mean on a discussion?
The sitting has ended. If you recorded what you decided, it says "decision recorded"; otherwise it simply closed with no decision.
This week
The week at a glance: what is slipping, which decisions wait on you, and tonight's close-out - what changed today and a rough plan for tomorrow.
What is this? Questions answered
What is the close-out?
The end-of-day record: what you did, what changed, tomorrow's rough plan. Saving it also tells her you have left for the day, which starts quiet hours.
Close-out
What changed today, with its receipts where they exist, and a rough plan for tomorrow.
Carbon ledger
What running her costs the atmosphere: a weekly report from estimated power use, priced in kg CO2-e with a factor you can inspect and correct.
What is this? Questions answered
Where does the factor come from?
It starts at 0.64 kg CO2-e per kWh - the Australian National Greenhouse Accounts Factors 2025 (DCCEEW), NSW scope 2 location-based value. It is yours to change below when a newer factor publishes; every change is receipted.
What does a report count?
The hours she runs, priced at the current factor. It is an honest estimate of her own footprint, not your household's.
Is anything sent away?
No. The ledger lives in her database; the report is built when you ask.
Factor
Reports
Alerts (ntfy)
Her line to your phone when something needs you - through the ntfy server you choose. Off until you switch it on.
What is this? Questions answered
Is anything sent right now?
Nothing, unless you switch it on below. Off means off: not even a test.
Who runs the server?
Whoever you point it at - self-host it or use ntfy.sh. The address and topic are stored here, never hard-coded.
What does a test do?
Sends one real notification through the configured topic so you can watch it arrive. That is all.
Status
Server and topic
Test
Self-healing
Her own health checks - database, schema, and the lanes she depends on - with repairs she is allowed to attempt and a log of everything she tried.
What is this? Questions answered
What does she repair on her own?
Only safe, reversible things - like clearing expired sessions. Anything riskier is reported, not attempted.
How do I know what she did?
Every check and every repair lands in the log below and in her receipts.
Does she hide failures?
No. A component that cannot check itself says so plainly instead of pretending to be healthy.
Log
Ecosystem
The other apps that make up her world, registered by you, checked by her - so "is everything up" is one look, not six tabs.
What is this? Questions answered
What does a check do?
One GET to the registered address, expecting a marker you chose. No logins, no crawling, no writes.
Can she add apps herself?
No. Only you register an app here; a duplicate name is refused.
What does stale mean?
The last check is older than her check rhythm, so the status may be out of date - run the checks to refresh.
Register an app
Phoenix backups
Her rise-from-the-ashes kit: real pg_dump archives, hashed at capture, verified on demand. Restore is not yet available.
What is this? Questions answered
When does she back up?
Sunday evenings Sydney time, once a week, plus any time you press the button.
How does a restore work?
Two steps: request a one-time token, then confirm with it within five minutes. The restore runs out of process - a fresh safety backup is taken first, other sessions are fenced off, and the archive hash and schema version are checked twice.
Why is restore marked disabled?
The repaired restore flow ships switched off until it passes independent audit. Backups and verify work today; the switch is hers to flip once the audit lands.
Status
Archives
Restore (not yet available)
Delegation (Baldrick)
Work she hands to Baldrick with your explicit yes: she classifies, he plans, you approve, the outcome is recorded so later delegations learn.
What is this? Questions answered
Does anything run without me?
No. A delegation waits for your approve, and external execution is currently switched off build-wide until the hardened approval flow passes audit - approvals stay waiting, nothing is sent.
What does "unknown" mean on a delegation?
He did not confirm the execute call, so he may still have run it. She never retries automatically. Use the reconciliation form only after you verify the exact request and evidence with him. It records your report, not an independent check.
What if his address changes?
An approval binds to the exact address the plan came from. If the address changes, the approval is invalidated and you ask for a fresh plan.
His address
Delegate a task
Delegations
Elaine · Almost Magic Tech Lab, Sydney · Private workspace for Mani - something wrong or missing? Say so in the Telegram chat and it lands on her floor.
R02 promise loop · R03 chronicle · R04 morning and news