Privacy notice
Last updated 6 September 2026
This notice covers jobmaxxing — the website you are reading, its waitlist, the hosted app, the public demo, and the Chrome extension. It is written in plain words on purpose. Where something is planned but not built, this notice says so rather than promising it.
Who we are
jobmaxxing is a product of Bountify, Inc. Bountify is the controller for everything described here.
- Post: [ADDRESS]
- Email: [CONTACT EMAIL]
- Waitlist and general contact: hello@jobmaxxing.ai
The website and the waitlist
- Your email address, if you join the waitlist. It is stored by Netlify (our form provider, United States) and used for one purpose: telling you when jobmaxxing is live. No newsletters, no other contact, and we never sell or share waitlist emails. We keep waitlist emails until the launch notice has been sent, then delete them within 90 days unless you have an account by then. When we invite you, the only thing that reaches our own systems is a one-way hash of your address — enough to recognise it when you sign up, not enough to reconstruct it or to build a second copy of the list.
- Anonymous visit counts via Umami, a cookieless analytics tool. It records page views and the referring channel, sets no cookies, and does not identify you.
This site stores one item in your browser — your light/dark theme choice — and nothing else.
The hosted app: what we hold
Your account runs as its own process with its own disk. Nobody else's account shares it. What lands on that disk falls into three kinds of thing.
- Your corpus — your own words. Your resumes, your profile sections, your stories, the companies, roles, fellowships, grants and ventures you track, your applications, and your application and interview prep. These are plain files. Automation never writes your resumes, your stories or your profile: the app writes them only when you click to save them yourself.
- Machine-written notes about your corpus. Research briefs, distillations, suggested entries waiting for you to accept or dismiss, and the analytics warehouse the board and the charts are drawn from. All of it is derived from your files and rebuildable from them.
- Records of the app working. An activity log with identifiers hashed rather than stored in the clear, a record of every AI call and what it cost, the transcripts of those calls, a step-by-step stream of the requests your app made, page-capture reports when a job page fails to read, and any masked session recordings the extension kept. The retention table below names each of these and how long it lives.
The Chrome extension
The extension reads job pages you are looking at so it can save them to your board. It never logs in for you and it never presses submit — those stay your hand. Its session recording is scoped to job-application hosts, masks every input by default, and blocks file inputs, password fields and any field labelled with EEOC, SSN or date-of-birth wording before any bytes exist. What is kept is masked page structure — tag names, an attribute allowlist and text lengths — plus the page URL.
How long we keep it
These are the windows from our own contract (SPEC-jobmaxxing.md, Data stewardship), reproduced word for word. They are enforced in code by a sweep that runs at boot and daily after that.
| store | window |
|---|---|
| trajectory snapshots | last 10 sessions or 30 days, whichever is smaller |
| trajectory step stream | gzip closed months; keep 12 months |
| LLM transcripts | 90 days; extension_parse transcripts are never hosted |
| recordings | user-managed; ~500MB/tenant, oldest-first eviction with warning |
| capture reports | 12 months |
More stores sit on your disk that the table above does not name. Their windows are stated here exactly as our data map states them.
| store | what it is | window |
|---|---|---|
| events | the activity log, identifiers hashed | not in D8 — kept |
| llm_usage | tokens and cost per AI call; no prompt text | not in D8 — kept for billing |
| resolutions | what the app guessed a page was, and your correction | not in D8 — kept |
| feedback | your own words about the product | not in D8 |
| capture_seen | the capture route's reply to each save | newest 2000 entries; 30 days |
| exports | a zip we built for you on request | until the actor's delete, or 30 days |
| observations | a daily freeze of your own ranked board | not in D8; kept, and deleted with your corpus |
| jobs.duckdb | the warehouse the charts read | rebuildable from your files |
Your corpus itself has no window. Your files stay until you delete them, or until your account is deleted.
Your consent choices
Two separate purposes, each one a choice you make and can change any day. Both start off. Refusing either keeps the whole product working.
- Share usage and error reports. Basic feature-use and failure reports to improve Jobmaxxing.
- Allow paid licensing of my activity. Uploading detailed activity to Bountify for preparation and paid licensing to AI research labs.
Local activity recording continues either way. These choices control sharing permission, not local recording. Detailed activity upload and preparation under the second choice serve the paid-licensing purpose; the first choice does not authorize sharing detailed activity.
Each decision and withdrawal is recorded with the choices, time and wording version (2026-09-purposes-v3). Earlier choices require fresh consent: both permissions stay off until you choose again. Your earlier record is preserved as history when you make that new choice. This history is included in your account activity download and removed when you delete your account activity data.
Rows recorded without consent remain local-only permanently. Earlier records are not upgraded to the new permission. Turning paid licensing off stops subsequent licensing exports, including rows recorded while it was on. Your personal activity download remains available either way.
Paid licensing is a commercial use of your data. Nothing has been licensed to anyone. Before third-party delivery, a redaction and synthesis layer is a hard precondition; that layer is not built. These choices do not by themselves activate an upload or licensing service. The extension's own emission settings decide what it sends to your app and cannot authorize either sharing purpose.
Other people's details
Hosted accounts do not store third-party contact records: the routes that would write a recruiter's or an interviewer's name are refused, the keys are left out on read, and a corpus that arrived carrying them is counted and then cleared by your own click, never rewritten by us. We do this because we cannot give a person we have never met the notice that Article 14 of the GDPR would require of us.
How the AI works
The AI is Claude, from Anthropic, reached through Anthropic's direct API or Claude Platform on Amazon Web Services under Bountify's account. When a feature runs, a relevant slice of your context is sent to the model and the answer comes back to you.
- Prompts and responses are kept as transcripts for 90 days, so you can read exactly what the model was told. Then they are removed.
- When the extension asks the server to parse a job page, that page text is never stored on a hosted account. The transcript row exists with its text fields empty and marked suppressed, because a loop detector counts rows for that surface.
- Token counts and costs are kept for billing.
We make no claim about what the model provider does with a prompt beyond answering it. If we can quote their contract on that, we will say so here, and not before.
Getting your data out, and getting it deleted
Ask us and we run two operations against your account: an export and a delete. Both walk every store listed above.
- The export is one zip: a file per store, plus a manifest with the row counts, the retention window per store, and the things the zip does not contain. A store holding nothing still gets an empty file, because "we hold nothing of yours here" is an answer.
- What the export does not contain. Your corpus is not copied into it as files — your resumes, stories, company and application markdown are yours on the disk, and they are exported by copying that disk rather than by duplicating the most sensitive text you have inside a log archive. The key that would un-hash the activity log is never in it either.
- The steps do carry your text. The trajectory step stream keeps the body of your own writes word for word, so a note you typed about a company, or a name inside an application answer, is in the export. Nothing is redacted, because an incomplete answer to "what do you hold about me" would be the worse failure. The manifest says the same thing.
- The delete removes your rows from every store. One store is deliberately excluded: the daily freeze of your own ranked board, which holds a hashed entity id, a score and a rank and no record about a person. It is deleted with your corpus. If one store fails, the failure is named and the rest still runs.
- Two identities. Your account has a server identity and one identity per browser where you installed the extension. "Delete everything about me" means running the walk for each, so tell us if you have used more than one browser.
Today both are self-serve: your Account tab carries an export button and a delete you confirm by typing, and they walk exactly the stores described above, while your corpus files stay yours to copy or delete as files. The extension's own options page already deletes the rows held for that browser.
For accounts with a shared weekly AI allowance, these buttons cover search activity. Separate customer-wide spending records hold your allowance, reset time and charges across searches, without prompt or response text. Deleting search activity does not reset that allowance or delete those spending records. Contact us for requests concerning the customer-wide records.
The public demo
The demo is open to anyone, has no accounts and no sign-in, and runs on a made-up dataset rather than anyone's real one. Its whole disk is wiped and rebuilt every six hours, so anything you type into it is gone within six hours. AI features and email capture are switched off there. Do not put real personal details into the demo.
Cookies
This marketing site sets no cookies at all — the analytics we use are cookieless, and the only thing stored in your browser is the theme choice. The hosted app sets one cookie: the sign-in session cookie minted by our load balancer, which lasts seven days. There are no advertising cookies and no analytics cookies anywhere.
We do not sell your data
We do not sell your data, and we do not share it with anyone other than the providers below who help us run the service. The one path by which anything of yours could ever reach a third party is the lab-licensing consent above: opt-in, revocable, blocked until the redaction layer exists, and never covering your resume, your stories or your files.
Who else touches it
- Amazon Web Services — hosting, storage, backups, and sign-in (Amazon Cognito).
- Anthropic — the Claude models, reached through Anthropic's direct API or Claude Platform on Amazon Web Services.
- Netlify — this marketing site and its waitlist form.
- Umami — cookieless visit counts for this marketing site.
Children
jobmaxxing is for adults. You must be 18 or older to hold an account.
Your rights
If you are in the UK or the EU, the UK GDPR and the GDPR give you the right to see the data we hold about you, to correct it, to have it deleted, to get a copy in a portable form, to object to or restrict some uses, and to withdraw consent at any time without giving a reason. Withdrawing consent does not undo what was lawful before you withdrew it. Our lawful bases are: contract, for running the service you pay for; consent, for the two purposes above; and legal obligation, for billing records. Ask us at [CONTACT EMAIL]. You can also complain to your data protection authority — in the UK that is the Information Commissioner's Office. Our controller entity, and any representative or data protection officer we are required to name, are [LEGAL CONTROLLER / REPRESENTATIVE — to be confirmed].
If you are a California resident, you may ask what personal information we collect, ask for a copy, ask us to delete it, and not be treated differently for asking. We do not sell or share personal information as the CCPA defines those terms. If lab licensing ever covered a California resident's data it would be a sale under the CCPA, it would need that person's opt-in first, and the notice and opt-out link would be here before it happened.
What is not built yet
- Edits in the app are not revertible — version history is the next thing we ship, and until it lands an edit is the edit.
- The redaction layer over trajectory steps does not exist, so no third-party export can happen.
Changes
If this notice changes, the date at the top changes with it. A change that widens how we use your data will be sent to you before it takes effect, and anything consent-based will be asked for again rather than assumed.