What does the "Recommended" sort do?
Left Join Jobs exists to make it easier to find tech careers at companies doing less harm, not just to filter by salary and location. "Recommended," the default sort on the search page, reflects that: it opens with a shuffled set of fresh postings, one per mission-driven company, before moving into a broader, diversity-aware mix across every company on the board.
More specifically: each company's own postings are first ranked by recency, so no single high-volume employer can flood the page with old and new listings alike. Within that, each mission-driven company's single most recent posting is promoted to the front, so the first several pages open with a shuffled set of fresh mission-driven listings. After that, the rest of the board follows the same company-diversity ordering, mission-driven or not. If you'd rather see strict chronological order instead, use the "Newest" or "Oldest" sort options.
The shuffle itself rotates once a day (at UTC midnight), not on every visit or every reload. That's deliberate, not a bug: page 1 and page 2 of the same search have to agree on ordering, or you'd see a posting twice, or skip one entirely, as you page through. A shuffle that re-randomized on every single request couldn't guarantee that. So reloading the page multiple times today shows the same order every time; check back tomorrow (or any later day) and it'll be a different mix. If it's been more than a day and the order genuinely hasn't changed at all, that's worth reporting. See below.
You do not need an account for any of that. The shuffle described above is the same for everyone, and nothing you click is recorded or used to reorder your results. If you're signed in and have saved a resume to your account, though, "Recommended" is personalized on top of the base ordering: postings whose listed skills overlap with your resume's detected skills are boosted ahead of the plain mission-driven tier (a posting that's both still wins over one that's only personalized). The same company-diversity rule applies: only each company's own single freshest posting is eligible for the boost, never an older one. You'll see a small "Personalized using your saved resume" note next to the sort control, and matching job cards get a "Matches your resume" badge, whenever this is active. Saving a resume is entirely opt-in (see the data section below) and only ever affects your own view, not what anyone else sees.
Sponsored placements from companies who want more visibility are still on the roadmap, not live yet; this FAQ will be updated when that ships.
What counts as "mission-driven"?
It's a separate thing from the values filters described below.
mission_driven is a single yes/no flag we set per
company (or, for government postings, per agency), not a
research finding about a specific practice like DEI or
government contracts. It's what "Recommended" sort boosts (see
above), and what the mission-driven filter checkbox uses.
There's no single rigid rule behind it. It's a curated set of categories we've built up over time: nonprofits and social-impact startups, progressive political and campaign technology organizations, labor unions, and government agencies. Federal jobs (pulled from USAJOBS) and city/state government jobs are mission-driven by default, with an explicit list of exceptions for agencies most people wouldn't consider mission-driven in this sense.
Two of those exceptions are whole departments, which means every agency underneath them is excluded too:
- Department of Homeland Security, and so every DHS component: CBP, ICE, TSA, CISA, FEMA, the Secret Service and the rest.
- Department of the Army and Department of the Air Force, and so every command under each, such as the Army Corps of Engineers and the Air Force Research Laboratory.
The rest are named individually, because the departments above them contain plenty of agencies nobody asked us to exclude. The Department of Justice, for instance, also covers the Civil Rights Division and the Antitrust Division, which still count as mission-driven here:
- Federal Bureau of Investigation
- Bureau of Alcohol, Tobacco, Firearms and Explosives
- National Security Agency
- Central Intelligence Agency
- Defense Information Systems Agency
- Defense Counterintelligence and Security Agency
- Naval Air Systems Command
- Drug Enforcement Administration
- Bureau of Prisons, also published as the Federal Prison System
- Department of War, the name the Department of Defense has used in its own communications since September 2025. Listed as that exact string rather than as "Department of Defense", so it doesn't sweep in every other part of the DoD.
Matching is on the posting's own agency and department names, so a sub-agency is caught by a department-level entry even when its own name looks nothing like the department's.
We'd rather be honest that this involves judgment calls than pretend it's a precise, formulaic system. If you think a company should or shouldn't carry this flag, the correction form works for this too. Tell us what you think we got wrong and why.
How are the values filters (Black leadership, DEI-friendly, no profits from ICE, etc.) sourced, and how confident should I be in them?
We research each company using public sources: news coverage, government records, campaign finance filings, and companies' own public statements. That research is updated from time to time rather than in real time. Treat each filter as a snapshot: a company's leadership, contracts, or public positions can change after we've checked.
Some listings show no values flags at all, and that is deliberate. Most postings here come straight from an employer's own careers page, and those employers we research one at a time. A smaller share arrive through aggregated feeds that carry many employers at once, and most of those employers have not been researched yet. Rather than guess, or borrow the flags from whichever feed a posting arrived on, we show nothing. A job with no badges is not a job that failed our checks. It is one nobody has checked yet, and we would rather say so than hand you a green tick we cannot stand behind.
Two limits worth knowing about the "not a known Trump donor" filter. First, it is mainly a check on the company itself: whether the business appears in federal filings as a donor to a Trump-aligned super PAC, which companies are allowed to give to. We also look for donations by named company leaders where we can, but that check is weaker, because people often write "self-employed" rather than their employer's name. So read a green tick as "this company is not a known Trump donor" rather than a statement about everyone who runs it. Second, the FEC only requires a committee to itemise a donor once their giving to it passes $200 in a two-year period. Smaller donations are real but appear only inside an unnamed total, so no public search can find them. A green tick means "no reportable contribution on record," not "never gave a dollar."
And one judgement call we make on purpose, so you are not surprised by it. We also count large Republican giving in a year Trump was running for president — 2016, 2020 or 2024 — even when the money went to the party rather than to a committee with his name on it. In those cycles the party's presidential campaign was his campaign, so a founder or chief executive who gave tens of thousands of dollars to the Republican National Committee in 2016 is, for this filter, a Trump donor. Giving in other years is treated differently: a big Republican donor in a cycle Trump was not the nominee is not marked here, and money to a rival's primary campaign is not either. It is a line we drew rather than one the filings draw for us, and reasonable people could put it elsewhere — so if a red mark here matters to you, the company's entry says which donations it rests on.
Our recommendation: pick one or two filters that matter most to you, rather than checking every box. Since an unchecked filter usually just means "no confirmed information," not a real "no" (see below), stacking many filters at once tends to hide real matches rather than surface better ones. If you're not sure where to start, we have the most confidence in the No known profits from ICE and No known defense contracts filters. Note the word "known": these come from searching federal award records, and those records have a real gap. Some defense work is done under what the Pentagon calls Other Transaction Agreements, and those do not appear in the public spending database at all. The government's own auditors have said the Defense Department does not track them well. Several large AI companies hold exactly this kind of agreement. So read a green tick here as "no federal contract turned up in the award records," which is good evidence, rather than as a guarantee that a company does no defense work. Where we know a company holds one of these agreements, we research it by hand instead. They're still the filters we trust most, because they're grounded in government contract-award records and reporting, which are generally more complete and verifiable than reporting on a company's internal leadership or political donations. The ICE filter covers more than contract records alone. See what that filter actually covers.
A checked box means we found a public source confirming it. An unchecked box just means we don't have confirmed information either way. It's not a "no." That distinction matters most for the leadership filters (Black leadership, women in leadership): we only ever check those when we can confirm them, and we never uncheck them as a negative. The absence of public information about someone's identity isn't evidence of anything, and we won't guess.
For the other filters, like government contracts or political donations, we occasionally note a claim as confirmed false when a public source says so directly, never as a guess.
We'd rather leave something unchecked than get it wrong. See something outdated or incorrect? Use the correction form to send us a public source and we'll take a look. And if there's a company missing from the board entirely, you can submit it here.
What does the "No known profits from ICE" filter actually cover?
This filter used to be called "No known ICE contracts," and it used to mean exactly that: does this company's own name appear on a contract with ICE, CBP, or DHS? That let a clean federal-database search hand a green check to a company profiting from immigration enforcement by every route except a direct contract, which is a false reassurance to exactly the person who chose this filter.
Both the rule and the name have since been widened, so that the filter asks the question you're actually asking: does this employer profit from immigration enforcement? A company fails this filter for any of three things:
- A direct contract with ICE, CBP, or DHS more broadly where the work serves immigration enforcement.
- Material financing of ICE contractors. A bank, asset manager, or investor providing significant loans, bond underwriting, or equity to companies like Palantir, GEO Group, CoreCivic, CACI, or L3Harris is profiting from immigration enforcement. A major investor in an ICE contractor is, for this filter's purposes, profiting from ICE.
- Sales funded by immigration-enforcement money, even one step removed. Think of a vendor whose software a sheriff's office buys with a state immigration-enforcement grant.
The evidence bar didn't move, only the definition. We still need a citable source naming that specific company; a general report that "banks finance private prisons" doesn't implicate any particular bank.
Two consequences worth knowing:
- A clean contracts-database search is no longer enough on its own to earn the check. We also have to have looked for financing and indirect entanglement and come up empty.
- Where we find entanglement that is real but doesn't fit cleanly, we leave the filter unset rather than checking it. So an unchecked box here sometimes means "we found something we couldn't cleanly categorise," not just "we haven't looked."
The No known defense contracts filter has deliberately not been widened the same way, which is why it still says "contracts" and this one no longer does. It means contracts, literally.
What does the "Not Musk/Thiel affiliated" filter mean?
This filter used to be narrower: "not a Musk company," meaning simply not owned or founded by Elon Musk. We broadened it to also cover Peter Thiel, since he's a similarly influential, similarly publicly-scrutinized figure in the same startup/venture-capital world as many companies on this board.
A company fails this filter (goes false) for either
of two reasons. (1) Musk or Thiel personally,
by name, currently, with a citable source, holds equity in or a
board seat at that specific company, most directly
Palantir, where Peter Thiel is a co-founder and
has served as Chairman of the Board since 2003 (per Palantir's
own investor-relations governance page); or (2)
the company is a portfolio company of Founders
Fund, the venture capital firm Peter Thiel co-founded
and owns. We treat a Founders Fund investment as a Thiel
ownership stake even when Thiel has no personal board seat at
that company, since Founders Fund's capital is, in a real sense,
Thiel's own money.
There is now a third way to fail it, and it runs in the opposite direction: a company that pays Musk or Thiel enough money to matter. The case that prompted this is a company with no Musk or Thiel investment at all. The full investor list was checked across every funding round and contains neither. But it has contracted to pay Musk's companies roughly $150 million a month for computing capacity through 2029. Under a strict ownership reading that company passes the filter. It shouldn't: someone who ticks this box to avoid enriching Musk would want to know their prospective employer is one of his largest customers.
So the standard is material economic entanglement, not only ownership. That is the same reasoning that made financing an ICE contractor count as profiting from ICE. The money can flow in either direction; what matters is whether the relationship is large enough to be a real fact about the employer.
As of this filter's last research pass, that combined standard is met by Palantir plus 20 companies cross-referenced against Founders Fund's own published portfolio page (foundersfund.com/portfolio): Airbnb, Anduril Industries, Affirm, Asana, Crusoe, DeepMind, Faire, Figma, Flexport, Lyft, Nubank, OpenAI, Persona, Ramp, Rippling, Scale AI, Spotify, Stripe, Twilio, and Zocdoc. None of Elon Musk's own companies currently have career listings on this board. Those are Tesla, Neuralink, The Boring Company, and SpaceX, which since February 2026 also contains xAI and X (formerly Twitter). Those three are one company now, under three names you may still see used for it.
One caveat worth flagging directly: Founders Fund's own portfolio page includes a few companies (like DeepMind, now wholly owned by Google/Alphabet since its 2014 acquisition) that Founders Fund may no longer hold current equity in. We didn't independently verify each company's present-day cap table; we're relying on Founders Fund's own page as the source of "who they consider part of their portfolio." If that page changes, this filter can go stale until our next research pass.
Same tri-state caveats as every other values filter apply here (see the sourcing question above): an unchecked box means "no verified affiliation found," not a guarantee that one doesn't exist, and this is a point-in-time snapshot that can go stale as company ownership, boards, and fund portfolios change. If you spot one we missed, use the correction form with a public source and we'll take a look.
What is the political alignment score, and how is it calculated?
A single number from −8 to +8 that rolls a company's values flags up into one figure, so you can sort or filter by it instead of ticking boxes one at a time. Higher means more aligned with the causes this board focuses on. It is not a rating of the company as an employer, and it says nothing about pay, management or whether you would enjoy working there.
It is arithmetic, not judgment. Nothing feeds it except the flags already shown on each listing. No extra research, no model, no hand-tuning per company. If you disagree with a score, the thing to argue with is one of the underlying flags, and the correction form is how to do that.
What each flag is worth
| Signal | If confirmed true | If confirmed false |
|---|---|---|
| Anarcho-syndicalist affiliated | +3 | n/a |
| Democratic-socialist affiliated | +2 | n/a |
| Democratic Party affiliated | +1 | n/a |
| Mission-driven organisation | +2 | n/a |
| Black leadership | +1 | n/a |
| Women in leadership | +1 | n/a |
| DEI-friendly | +1 | n/a |
| LGBTQIA+ friendly | +1 | n/a |
| No known profits from ICE | +2 | −2 |
| Not a known Trump donor | +2 | −2 |
| No known defense contracts | +1 | −1 |
| Not Musk/Thiel affiliated | +1 | −1 |
| Not primarily an AI company | +1 | −1 |
| No known crypto leadership | +1 | −1 |
A dash means that flag can never subtract. Add up every column that applies, then clamp the result to the −8 to +8 range.
Why some flags can go negative and others can't
The eight signals in the top half of the table are true-only. We will confirm that a company has Black leadership; we will never assert that it doesn't. Nobody publishes a press release about the executives they don't have, so silence there is an absence of information, not a finding, and recording it as a negative would be inventing a fact. An unticked box in that half means "not confirmed," full stop.
The six in the bottom half check something you can actually look up: a federal contracts database, a donation record, a company's own description of its business. A real search that comes back clean is itself evidence, so those can move the score in either direction: up when we searched and found nothing, down when we searched and found something.
Two rules that stop the score being gamed
Zero usually means "we don't know", but not always. A company with nothing confirmed either way scores 0, and most companies on this board sit at or near 0 simply because nobody has researched them yet. A 0 and a −2 are very different claims; a 0 and a +6 may be too.
The honest caveat: 0 is also where a thoroughly researched company lands when its positives and negatives happen to cancel. One company on this board scores 0 while carrying three confirmed adverse findings. It sells AI, holds a defense agreement, and is heavily entangled with Elon Musk. A clean ICE record and clean crypto record offset them. So a 0 is a prompt to open the listing and read the individual badges, not a verdict. We'd rather tell you this than let the number quietly imply more than it knows.
A clean record alone tops out at +6. The six checkable flags add up to +8 on their own, which would let a company with no confirmed positive signal whatsoever sit at the very top of the scale purely because we went looking for adverse findings and didn't happen to find any. That is a statement about our research, not about the company. So if a company has no confirmed identity flag, is not mission-driven, and has no political affiliation on file, its score is capped at +6. Reaching +7 or +8 takes at least one confirmed positive signal on top of the clean record.
A worked example
A worker-owned software co-op with a woman co-founder, confirmed clean on all six checkable flags, and no party affiliation on file: +2 (mission-driven) +1 (women in leadership) +2 (no ICE contracts) +2 (not a Trump donor) +1 +1 +1 +1 (the remaining four) = +11, clamped to +8. The cap does not apply, because it has two confirmed positive signals.
A defense-sector AI startup founded with Founders Fund money, with a published diversity policy: +1 (DEI-friendly) −1 (defense contracts) −1 (is an AI company) −1 (Musk/Thiel affiliated) +2 (no profits from ICE) +1 (no crypto leadership) = +1, with "not a known Trump donor" unresearched and therefore worth 0.
Treat the number as a rough, best-effort summary. It is a starting point for your own reading about an employer, not a substitute for it.
What do the anarcho-syndicalist / democratic-socialist / Democratic Party filters mean?
These three were added at a user's explicit request, specifically to feed the alignment score above with a more fine-grained political signal than "mission-driven" alone provides. Like the Black leadership/women-in-leadership filters, they're true-only: we never mark an org "not affiliated" as a confirmed negative, since the absence of a public affiliation record isn't evidence that none exists. An unchecked box just means no formal affiliation was found. It is not a claim that the org is apolitical or opposed to that movement.
The evidence bar is deliberately narrow and formal for the anarcho-syndicalist and democratic-socialist flags: a general political vibe or a founder's personal voting record doesn't qualify. A box is only checked when either (a) the org is itself a formal chapter, branch, or party-administered platform of that movement, (b) a current named leader holds publicly documented, current formal membership or office in it, or (c) a single named exception, decided deliberately and applied to exactly one organisation, where the org's own platform and electoral strategy is substantively democratic-socialist without formal membership, backed by strong, specific, on-record characterization from a leading democratic-socialist figure. This third path is not a general standard; it does not automatically extend to other minor left parties without their own similarly strong, named citation.
The Democratic Party flag started on that same narrow standard in its first research pass, but was widened twice at the user's explicit request in follow-up rounds. The first widening added orgs that function as de facto Democratic Party electoral infrastructure: the party's own fundraising platform, a vendor built to serve the party and Democratic campaigns as its core purpose, a PAC or legal/media operation whose stated mission is electing Democrats, or a minor party that runs in formal cross-endorsement coalition with the Democrats. The second widening added political consulting firms: digital, data, or communications agencies whose own public self-description names Democratic and progressive campaigns and candidates as their core client base. Under both widenings, a firm still doesn't qualify just because some of its clients lean Democratic; its own stated mission and client focus has to be Democratic/progressive campaigns and candidates specifically, not a mixed partisan/nonpartisan roster. A leader simply voting Democratic or donating to Democratic candidates still doesn't qualify on its own either; that's what the "not a Trump donor" filter already covers.
Under that standard, thirteen orgs on this board currently clear the Democratic Party bar. Formal/infrastructure matches: NGP VAN (whose VoteBuilder/VAN platform is the DNC's own administered voter file, the de facto sole DNC voter-file vendor since 2004), ActBlue (the Democratic Party's own official small-dollar fundraising platform, used near-exclusively by Democratic candidates and committees), EMILY's List (a PAC whose entire mission is electing Democratic women), Catalist (a data/analytics vendor built to serve Democratic campaigns and progressive organizations as its core purpose), Democracy Docket (a voting-rights legal/media outlet founded and led by the DNC's own outside counsel), and the Working Families Party (which runs in formal cross-endorsement/ fusion-voting coalition with Democratic candidates in the states where it's active). Campaign-consulting matches: Blue State Digital (its dedicated political practice works exclusively with Democratic and progressive campaigns, with clients including the DNC and Barack Obama's presidential campaigns, though the broader agency also serves some non-political corporate and nonprofit clients), Middle Seat (a progressive digital fundraising and media firm founded by former Bernie Sanders 2016 campaign staff, working exclusively with progressive causes and candidates), BerlinRosen (a PR/communications firm working primarily for Democratic campaigns, progressive nonprofits, and labor unions), Civis Analytics (a data-science firm founded by Obama 2012's analytics director that serves Democratic campaigns exclusively on the political side of its business, though it has expanded into non-political government and corporate work), Grassroots Analytics (a donor-data firm described in its own reporting as a "Democratic tech firm" serving progressive and Democratic causes), and Movement Labs (a progressive organizing-tech consultancy working with progressive campaigns and organizations nationwide). None of these thirteen holds a formal officer title inside the Democratic Party's own organizational chart, which is why an earlier, narrower research pass on this board left them unchecked. Under the widened standard now in effect, their own stated mission and client focus qualify them.
On the democratic-socialist side, exactly one org clears the bar: the Working Families Party. WFP doesn't formally identify as socialist or hold DSA membership, but its platform and electoral strategy are close enough that Sen. Bernie Sanders has called it "the closest thing there is to a political party that believes in my vision of democratic socialism", a strong, specific, on-record characterization from arguably the most prominent democratic socialist in US politics. No other org on this board clears either the anarcho-syndicalist or the (unwidened, formal-membership) democratic-socialist bar, and that remains the honest, correct result for the rest of the board, not a sign the research was incomplete.
Same caveats as every other values filter: this is a point-in-time snapshot, not a guarantee, and we'd rather leave something unchecked than get it wrong. If you know of an org on this board that should clear this bar, use the correction form with a public source.
What data do you keep on me?
We never share or sell your data. Nothing described here is given to advertisers, data brokers, or any other third party. The account data below exists only to run the features you've opted into (signing in, saved searches, resume matching), full stop.
Creating an account only ever asks for an email address, which is what a magic-link sign-in needs, and it's all we store on your account (along with whether you're on the free or subscriber plan, and whether you've unsubscribed from email alerts). We don't ask for or store a phone number, mailing address, or your name.
If you subscribe, your card details go to Stripe and never touch our servers — there is no card number, expiry or security code anywhere in our database, which is why removing your card is something we ask Stripe to do rather than a field we clear. What we store is two Stripe reference ids, so we know which subscription belongs to which account. Your email address is shared with Stripe so they can send you receipts. You can cancel, and remove your saved card, from your profile page.
Saved searches store the filters you searched with (keywords, tags, location text, etc.) and, if you leave email alerts on, when we last checked for new matches so we don't email you the same posting twice. Delete a saved search any time from your profile page and it's gone.
Resume matching is opt-in on top of opt-in: by default, uploading a resume on the resume matches page reads it, finds your matches, and discards it. Nothing is saved. You get two different ways to save something for next time, and neither is required: "save this resume" stores the extracted TEXT of your resume (not the original PDF/DOCX file itself), while "save just my skills" is a lighter option that keeps only the list of skills we detected, not the resume text, so anything else your resume happens to contain (a phone number, address, past employer names, whatever) is never written to our database in that mode. If you do save the full resume text, whatever your resume happens to list is part of that text, so it's on you to leave that off a resume you choose to save here if you'd rather we not have it; we don't ask for, extract, or use that information separately from the resume text itself. Remove either kind of saved resume data any time from the resume match page or your profile page, and it's deleted immediately. There is no automatic expiry otherwise.
Skills to highlight (on your profile page) is a separate, simpler option: pick skills directly from a checklist to personalize your search, with no resume involved at all. It's just a list of skill names you chose. Nothing else about you is inferred or stored from it.
Job clicks (which postings get clicked through to the employer's own site) are logged for internal analytics on which listings actually get engagement. The row is two things: which job, and when. Not who — every click is recorded the same way whether you're signed in or not, so these counts can tell us a posting is popular but can't tell us you looked at it. And nothing about what happens after the click reaches us: once you're on the employer's site you're on their site, and whether you read the posting or apply is between you and them.
We set no cookies. Not one — no advertising cookies, no analytics cookies, no third-party pixels, nothing that follows you anywhere. There is no cookie banner on this site because there is nothing to ask you about, which is the version of "we respect your privacy" we'd rather ship than a pop-up.
That took a deliberate decision, so here it is. We briefly built a cookie that tied your job clicks to your account, and then took it out. The only thing it would have fed is a recommendations feature that doesn't exist yet, and keeping it would have meant asking every visitor for cookie consent. Not collecting it was the better trade. Job clicks are still counted — just as counts, with nothing identifying you.
Before that, the click link carried your sign-in token in the web address itself, where it ended up in your browser history and in our server logs. That's fixed. But if you clicked job links here before September 2026, those old history entries are still on your own computer until you clear them. We'd rather tell you than leave it out.
One exception, and only if you pay us.
Stripe
handles our payments, and Stripe's script sets a couple of
cookies of its own (__stripe_mid and
__stripe_sid) for fraud checking. Its script only
loads at the moment a card form actually appears in front of you,
so if you never enter a card, it never runs and those cookies
never exist. They're Stripe's, on Stripe's terms, and nothing
else on this site sets anything.
What we do keep in your browser isn't cookies but local storage, which stays on your device and is never sent anywhere unless you're the one asking for something: your sign-in token, the email address we greet you by in the header, a list of the job postings you've clicked through to (that's what puts the grey "Viewed" marker on a card), and, while a resume match is running, a reference to that match so the progress indicator survives a page reload. All of it exists to run something you asked for — that's why none of it needs a consent prompt either.
The "Viewed" list in particular never leaves your device. We deliberately didn't build it on the server, where it would have meant keeping a record of every posting each account had looked at. The trade-off is that it's per-browser: it won't follow you from your laptop to your phone, and it resets if you clear your browsing data. That's the cost of us not holding it, and we think it's the right way round. Signing out clears your sign-in token and cached email; clearing your browser data clears everything above, the viewed list included.
The correction, submit-a-company, and list-your-job forms ask for an email address so we can follow up, but that address is only ever used to send us a one-time notification. It is not written to our database at all.
