Second of three posts on how a forum full of anonymous strangers works out who to trust. Part 1 covered the signals: the slang, the tags, the way a member's posting history gets read, and the slow public grind a provider goes through before anyone calls it safe. This part covers what happens when that goes wrong, somebody posts an accusation, and several thousand people have to decide what to believe.

A note on names: @handle is a real forum account, quoted from a public thread. Companies appear without the @. LES is LowEndSpirit, where this is posted; OGF is LowEndTalk, the other forum, and I say so when a quote comes from there.

In this part
1. How a scam alert actually spreads
2. The deadpool list as a trust ledger
3. Second-order trust, or who vouched for the voucher
4. There is no verified badge, and then in 2026 there nearly was

How a scam alert actually spreads

One angry customer posting "this provider is a scam" does almost nothing on its own. That is not cynicism about complainants, it is just what a decade of watching these threads teaches you. Some accusations are right, some are a misunderstanding about a billing cycle, and from the first post you frequently cannot tell which. So the community does not treat the accusation as the verdict. It treats it as the opening of a hearing, and three things decide the outcome: what is in the thread, whether anyone independently corroborates it, and, weighted far above the other two, how the provider answers.

Two real threads show the whole range.

In August 2024 a member called @xbugx posted about Dedi Rock. Their service had been cancelled on the 24th, days after being switched to a six-month billing period, with no refund. The post did not hedge: "I do NOT recommend Dedi Rock. They are an ultimate scam, and their support is beyond terrible."

Note what happened next, because it is not what an outsider expects. The forum went after the accuser first. @tetech asked, bluntly, "Where's the ticket screenshots? Given they replied next business day and you've had a meltdown, the issue must have been urgent." Then, more specifically: "Where is the screenshot from Ticket ID: #MRB-399189?" @bikegremlin asked the question that actually decided it: "When they switched you to the 6-month billing period and issued an invoice for it (on the 19th), did you pay that invoice before the 24th (when your service was cancelled)?"

Then Dedi Rock showed up:

"Hey, very sorry about this. There was definitely an error however nothing was done manually on this end. It appears it was automatic. We are getting to the bottom of it now to make it right (we have never seen this error before). If you should choose to continue with us I will compensate you a free month or two for the trouble, if not I will process the refund in full."

Dedi Rock, the provider, replying in the thread

Apology in the first four words. A specific mechanism. An admission of not knowing the cause yet. Two concrete remedies, one of which costs them the customer and the money. The thread title was then edited to "Dedi Rock payment processing hiccup - resolved," which is the forum's permanent public verdict on that accusation, sitting in the URL forever.

Now the other direction. In late January 2024 Limewave went dark, the whole operation, AS36369, gone. The operator, @Dvo, eventually posted at length over on OGF: the shutdown had not been planned before the end of Q4, the company was profitable but had missed targets, 62% of customers were paying late, nodes would come back through OVH for about three weeks so people could pull their data. On exit scams he wrote, "The problem with exit scams is that fortunately, they don't exist anymore." On chargebacks: "If people think paying $12/yr for a service, using the service for 6 months, then charging back the full $12 when they're only lawfully entitled to a $6 refund, I'm not going to sit there and fuck around with arguing with people."

It was thousands of words. The community's response to all of it fit in one line, from @MrFd:

"Man, @Dvo write thousands word, but not a single 'sorry'?"

@MrFd, LES member, in the Limewave thread

@treesmokah put the reading plainly: "Funny, basically exit-scamming but saying he isn't and deflecting." Members went to their banks. @Blohsh, who had a year prepaid and around $250 in credit, was one of many.

Same forum, same mechanism, opposite outcomes, and the length of the response had nothing to do with it. Dedi Rock wrote four sentences and kept its reputation. Limewave wrote an essay and lost everything. What separates them is that one of them accepted the cost and the other explained why the cost was somebody else's fault, and that distinction is genuinely hard to fake under pressure in public, which is exactly why the community weights it so heavily.

There is a rule layer under all this now. Since June 2026, a negative review posted on LowEndSpirit without proof goes to the Rants section, which is de-indexed. The complaint still exists, members can still read it, and it stops appearing in Google against the provider's name. That is a sharper piece of design than it looks: it separates "I want to warn people" from "I want to permanently damage a company on the strength of my say-so," without requiring a moderator to decide who is telling the truth.

The deadpool list as a trust ledger

Nobody is assigned to maintain it. It updates itself every time somebody gets burned.

Institutions remember through records, audits and filings. This community remembers through one enormous thread. The "2015 Deadpool List" on LowEndTalk ran up more than 13,000 views and 265 replies, and it does the job a credit bureau does, by completely different means: it aggregates dispersed private experience into one checkable public record.

Read as a trust document rather than a history, what it answers is a question no single member could answer alone. Has this company already burned people, and how. Somebody evaluating an unfamiliar provider can search that thread and inherit a decade of collective memory they were not around for. BigVPS.io, Viral VPS, Haphost, BlueVM, GoodHosting, Heroicvps, onebesthost.com, all sitting there in public.

The record is more careful than a blacklist would be, which is what makes it usable. GreenValueHost was salvaged by XFuse/TacVPS and later absorbed by Hostress. NatVPS and 32mb.club went to MyServerPlanet. Reversehost's clients ended up at HostUS. Peak Servers was bought outright by ServerHub. Neximweb just discontinued its VPS plans and carried on doing something else. "Acquired" and "vanished with your data" are different outcomes, and the thread distinguishes them, because the people writing it were the people affected and they cared about the difference.

The single most damning entry in the whole ledger is not about a provider at all. In December 2019, twenty separately branded budget VPS companies, sharing page structure, copied marketing text and the same CAPTCHA implementation, sent an identical shutdown notice within hours of each other and went dark. Seventeen of the eighteen hosts advertised on the community's own deal-listing site in the preceding two months were among the twenty. The record kept by the community indicted the community's own promotional channel, in public, and stayed there.

What makes this ledger durable is that nobody maintains it. There is no custodian, no assigned role, no funding. It updates passively, every time somebody who just got burned writes down what happened, mostly so the next person does not have to find out the same way. That is a different model of institutional memory than any formal archive, and its running cost is zero.

Second-order trust, or who vouched for the voucher

Almost nobody verifies a provider directly. They trust somebody who did.

Here is the part that undercuts the tidy version of this story: most trust here is not first-hand at all. Very few members personally evaluate a provider. They trust somebody whose judgment has been reliable before, who trusted somebody else, and so on. That is second-order trust, and the system runs on it far more than on independent verification.

This is unavoidable and mostly fine. Nobody has time to research every provider mentioned here or read every complaint thread to the end. Second-order trust compresses years of distributed evaluation into "people whose judgment I have tested seem comfortable with this," which lets one person act on far more information than one person could ever hold.

The clearest artifact of it is this forum's own FAQ. A new member called @Joseph asked, in a thread, what LES was and where it came from. @bikegremlin, a member with no staff role at the time, wrote an 8-minute article answering it on his personal site. That article was then republished, with permission, as the forum's official FAQ. Nobody on the admin team wrote the document that explains this place to newcomers. A volunteer did, because somebody asked, and the community adopted it. The same logic put him in a moderator role in 2022, after @Mason and I stopped appointing people and asked the membership instead, in a public thread that attracted several joke bitcoin bribes alongside the real nominations.

The most expensive version of this I have watched involved no forum thread at all. When NexusBytes was dying, its owner @seriesn seriously ill and its infrastructure gradually running out of money, @jarland, who runs the competing email host MXRoute, quietly funded its continued operation for months so that a rival's customers had time to migrate their data out. He was under no obligation and got no forum tag for it. What he got was a reputation that now travels ahead of him, which is the whole currency this system runs on.

The failure mode is exactly what you would expect. Errors propagate rather than staying local. A member with ten years of good calls who vouches for the wrong provider does more damage than a stranger making the identical claim, precisely because the track record is what made the recommendation worth anything. There is no mechanism to catch that in advance. There is only the slow public correction from the last section, where somebody eventually posts a contradicting experience and the chain re-evaluates itself.

What keeps it from collapsing into an echo chamber is the number of chains running in parallel. A recommendation that traces back to several unconnected members with their own separate track records is worth substantially more than one that traces back, on inspection, to a single enthusiastic source. If you are reading this community from the outside, that is the most useful habit to steal: before weighing a recommendation, work out how many independent people it actually comes from.

There is no verified badge, and then in 2026 there nearly was

Nobody certifies your judgment here. The entry gate, though, just got a vote attached.

It would be tidier to end this with a description of the verifier role: some recognised status, earned through a defined process, marking certain members as authoritative. That would be false. There is no such title. Nothing on this forum certifies that a given member's judgment is sound, and no tag means "believe this person." OG marks tenure. Content Writer marks contribution. Hosting marks a commercial interest. None of them is an opinion about whether you are right.

That absence has a real advantage, and it is worth naming because it is the reason not to fix it. A formal verifier tier is a single point of capture. Whoever decides who gets verified decides whose word carries official weight, which rebuilds the exact centralised authority that this community was founded to get away from. Distributed credibility is slower and harder to explain, and it cannot be captured by controlling one role.

That is the principled version. Here is what actually happened in 2026.

In June, provider vetting became formal: a new provider now applies through the registered-providers page, and a public thread is opened so members can weigh in before the tag is granted, with automated voting behind it. Then, mid-year, the same treatment came for members. New registrations answer a question with a minimum answer length, in one shared new-members thread, and the community votes. At -7 they are auto-banned. At +7 they are promoted to standard member. Until promoted, they cannot see offers or domain sales and cannot start threads. Only members with at least 25 karma can vote, a threshold Anthony Smith (@AnthonySmith) said had been discussed at 100 and deliberately left low to see how it went.

The numbers he reported on 11 August were stark: new memberships down 66% over thirty days, new members posting in topics up 141%. Bots filtered, real people participating faster.

His own framing of the change was the interesting part:

"the previous system was 1 question and 1 person deciding, it's just more visible now."

Anthony Smith (@AnthonySmith), LES founder and admin, on the new member verification system

The argument that followed is the thing worth reading, because the community immediately found every failure mode a formal system introduces. @somik argued that anonymous downvoting has no accountability attached and discourages legitimate signups. @treesmokah, after a new member was downvoted for wanting hosting for something politically questionable, argued that community voting produces emotional judgment where neutral staff judgment used to sit, and compared it unfavourably to Reddit. @ahnlak countered that vague phrasing in a hosting request is itself a signal. @MichaelCee resolved it against the house rule: if you're racist, you're a dick. @Lee complained the new-member threads were cluttering the homepage. I proposed raising the promotion threshold to reduce how volatile the voting is.

Notice what got formalised and what did not. The community built a gate, with a number attached, for entry. It still has not built anything that certifies judgment, and nobody in that argument proposed one. You can now be voted into the room. Nothing votes on whether you are worth listening to once you are in it, and that distinction is the load-bearing one.

Part 3 is about where all of this breaks: the pile-ons, the unfair outcomes, and what actually transfers to a community that is not this one.

**Missed part 1? It is available here. **

Someone is about to send fifteen dollars to a company that has existed for four months, run by a person whose real name they will never learn, for a server in a datacenter they cannot name. They do not check a business registry. They do not read the terms of service. They post six words in a thread, "anyone actually used these guys?", and then they act on whatever three strangers with cartoon avatars tell them.

That is the transaction this whole community runs on, thousands of times over, and the strange part is that it mostly works. I have been a co-admin at LowEndSpirit since February 2021, which means I get to watch the machinery from the inside, including the parts that are ugly. This is the first of three posts about how that machinery actually works: what the signals are, how they get read, and where they fail.

A note on names: anything written as @handle is a real forum account, and every quote attached to one comes from a public thread you can go and read. Providers and companies appear without the @, so @Mason is a person and Dedi Rock is a company.

In this part
1. Why strangers trust strangers here
2. The vocabulary of trust
3. Reading a new member's post history
4. How a provider earns credibility

Why strangers trust strangers here

Every mechanism you would normally lean on to check a company is missing at this price point, and it is missing for structural reasons rather than accidental ones. Nobody runs a credit check before selling a three-dollar-a-month VPS, because the margin on that sale will not cover the cost of running one. Nobody demands proof of a business address. There is no ratings agency covering the low end of hosting, no ombudsman, and if you file a chargeback for eighteen dollars your bank will process it long after your data is gone.

The barrier to launching a hosting company is a website, a payment processor, and a few servers rented from somebody bigger. Which means the barrier to a hosting company disappearing is exactly as low.

So the market that grew here had a choice: operate blind, or build a substitute for institutional trust out of the only material available, which was people talking to each other in public, repeatedly, over years. It built the substitute. Not deliberately, and not by design, but the thing that exists now is a working trust system with recognisable parts: a shared vocabulary, a set of legible behavioural signals, a public dispute process, and a written record of everyone who has already burned somebody.

What it does not have is a reputation score. LowEndTalk and LowEndSpirit both run on forum software that could easily display one, and neither does anything meaningful with it. There is no number next to a username that tells you how much to believe them. There is no verified badge for a provider, no star rating, no aggregated feedback percentage of the kind eBay built its entire marketplace on. Nothing tells you who to trust. You are expected to work it out.

The interesting question is how thousands of people manage to work it out anyway, consistently enough that the market functions. The answer is not that everyone here is unusually shrewd. It is that the signals are real, they are cheap to read once you know what to look at, and most of them are expensive to fake. That combination is what makes an informal trust system hold up under pressure, and the rest of this post is about the specific shape those signals take here.

I should say up front what this is not. It is not a claim that this system is better than a well-run institution. It is a claim that it is what you get when no institution shows up at all, and that it works better than it has any right to.

The vocabulary of trust

Small chicken, mother chicken. The slang is a map of what the community actually tracks.

Communities coin words for the things they need to talk about constantly. If you want to know what a group actually pays attention to, read its slang, because nobody invents shorthand for a concept they only need once a year.

The most visible piece of vocabulary in this corner of the internet is MJJ, which turns up constantly on both forums and confuses almost every outsider who tries to research it. @Mason, who ran this place alongside me for years, has the short version:

"MJJ is a short phrase 没鸡鸡 in Chinese hosting forums. Direct translation means: Lack of VPS. Meanwhile the homophonic is: Lack of penis. The word used by poor members self-deprecating. Targets who have no money but wish to buy a good server."

@Mason, LES former admin

The wider Chinese-language VPS scene, mainly Hostloc and NodeSeek, gets there by a second route that lands on the same three letters. A VPS is a 小鸡, a "small chicken." A dedicated server is a 母鸡, a "mother chicken." People who buy them are 买鸡鸡的人, chicken-buyers, which shortens to MJJ as well. Both readings describe the same person: somebody with no real money who wants a good server anyway.

The rest of that vocabulary follows the same logic. 杜甫 is the Tang dynasty poet Du Fu, and also what you get when a Chinese input method mangles 独服, "dedicated server," so the poet's name became the word. 吃灰 means "eating dust," which is what your VPS does after you bought it in a panic during a flash sale and then never found a use for it. 传家宝 is a "family heirloom," a plan so cheap and so long discontinued that you will never cancel it, because the moment you do it is gone forever.

One note for anyone researching a community from the outside: a plain web search got the MJJ etymology roughly right before I ever asked @Mason, and the two matched. Outside research is not automatically wrong. It is just unverified until somebody inside confirms it, and those are different states even when the answer turns out identical.

LowEndSpirit has grown its own vocabulary on top of the imported stuff, and the most telling entry is not a compliment. LowEndTalk is routinely called OGF, the "Other Green Forum," a name used on both sites so nobody has to type the real one. WHT is WebHostingTalk, generally invoked as the cautionary example of moderation so heavy that helpful posts get people banned. When @bikegremlin, a moderator here at LES, linked to LowEndTalk in the article that later became this forum's own FAQ, he added: "Yeah, it's a nofollow link, I don't vouch for that site." A rel="nofollow" attribute is a trust statement now. That is how granular this gets.

The rest of the local vocabulary is tags. OG sits next to the usernames of members who came over from LowEndTalk during the 2019 migration, a marker of tenure that confers no privileges at all. Content Writer goes to people who write guides for the LES Talk section. Hosting and Service Provider mark commercial accounts. Regulars call themselves LESbians, which tells you roughly what to expect from the place. And deadpool, the word for a provider vanishing, gets used as a verb without explanation.
_At the time of writing, the Content Writer tag was still available; it is no longer.
Some other tags mentioned in this series are also removed, or might be in the future. _

Notice what the tags actually track: how long you have been here, what you have given away for free, and whether you are trying to sell something. Those are three of the four things this community weighs when deciding whether to believe you. The fourth is your track record, and no tag exists for it.

There is a joke version of all this, and it is load-bearing. @bikegremlin also wrote a satirical "New proposed LES FAQ" in which OG stands for "Openly Gay," moderation is fictional, the forum secretly runs on an Arduino in another member's bathroom, and membership costs a kidney. It is funny only if you already know the real rules, which makes it an unusually efficient in-group test. Nobody wrote that as a trust mechanism. It works as one anyway.

Reading a new member's post history

The signal is in the pattern, not in any single post.

Ask me how I decide whether to believe a member I have never interacted with, and I cannot give you a checklist, because I do not use one. Neither does anyone else here (except for the rare few). What actually happens is closer to pattern recognition, and the patterns are specific enough to write down even though nobody normally bothers.

The strongest single signal is unprompted helpfulness in threads with nothing in them for you. A member who turns up in somebody else's WireGuard config thread, reads the paste, spots the missing AllowedIPs line and says so, has just demonstrated something no sales pitch can. They spent twenty minutes on a stranger's problem in a thread that will never make them a cent. Promotional accounts almost never do this, because there is no return on it, which is precisely why its presence or absence separates them so cleanly.

Writing style matters too, in a way that is hard to formalise and easy to recognise once you have read enough of both kinds. The tell is specificity, and it works in both directions. A real complaint arrives with dates, ticket numbers, and the exact wording of the error. In one dispute I will come back to in part 2, the provider's own reply quoted ticket #MRB-399189, and that number did more to settle the argument than any amount of tone. A fabricated complaint tends to be furious and vague at the same time. Genuine praise usually comes with a caveat attached, because people who actually use a service have noticed at least one thing they dislike about it. Marketing copy never has a caveat, because marketing copy is not describing an experience.

Tenure adds weight without being sufficient on its own. A five-year-old account that has only ever posted offers reads very differently from a five-year-old account with a scattered history of ordinary participation. What tenure buys is not credibility directly. It buys a longer record for other people to have already evaluated, which is a different thing, and it is why a well-written post from a two-week-old account gets read with more suspicion than a badly written one from a member who has been arguing about NVMe pricing since 2020.

None of this is unfoolable. A patient bad actor with a year to spare can build an account that passes every one of these checks, and occasionally one does. What the signals buy is cost. Faking a year of unpaid helpfulness across dozens of unrelated threads is genuinely expensive in the only currency that matters here, which is time, and most scams are not run by patient people.

How a provider earns credibility

There is no fast lane. The only route to being known good is being watched for a long time.

A hosting provider showing up here starts at zero, and there is no amount of money that changes that. What there is instead is a gate, and then a long wait.

The gate is procedural. Under the forum rules that took effect on 11 June 2026, a new provider applies through the registered-providers page, and a public thread is opened so the community can look at them before the tag is granted. The standard, in the admin's own phrasing, is whether "they want to be here, and the community receives them ok." Post a commercial offer without that tag, and you get banned, with narrow exceptions for domain sales, service transfers, and things given away free.

That rule does not make anybody trustworthy. It does one specific job: it stops a provider from building fake enthusiasm from an account that looks like an ordinary happy customer. Whatever credibility a provider earns here, it gets earned in the open, as a known provider, with the commercial interest stated. That is the entire point.

The same rules go further than tagging. Providers may post once per seven days in each category. Bumping your own thread is not allowed unless you have something genuinely new to add or somebody has commented in the last day. You may not require people to comment in your thread to earn a bandwidth upgrade, which kills the oldest trick for manufacturing the appearance of an engaged customer base. Affiliate links must be prefixed with "(aff)", with the rule stated as "abuse it and lose it (ZERO TOLERANCE)". Every one of those is aimed at the same failure: a provider buying the appearance of a reputation instead of earning one.

What the rules deliberately do not do is guarantee anything. LES describes itself as a buyer-beware market and means it. For most of its life it wrote strict price caps into the rules to keep every offer genuinely budget-friendly; as of June 2026 those limits are suspended entirely for a six-month trial, so even the old proxy of "if it is on LES, it is cheap" no longer holds. The forum removes obvious fraud and known fraudsters, and leaves the rest to you.

Which leaves time as the only real mechanism. A provider that has run cleanly for three years has something no advertising budget can buy: a searchable history of customers vouching for it in threads it did not write, on a forum where it cannot delete criticism. That is distributed corroboration from people with no stake in its success, and it is hard to fake at scale for exactly as long as it takes to accumulate.

The best evidence of what a provider is actually made of tends to arrive at the worst moment. When DataIdeas closed in June 2025, the founder gave real notice, refunded the unused portion of prepaid plans, and offered dedicated and colocation customers the chance to buy their own hardware outright rather than watch it get repossessed. He was going out of business, with nothing left to gain from any of it. That is what a reputation is actually for, and part 2 is about what happens when it goes the other way.

Last of three posts on how a forum full of anonymous strangers decides who to trust. Part 1 covered the signals people read. Part 2 covered what happens when somebody shouts scam, and the 2026 arrival of an actual vote at the door. This part is the uncomfortable one: where this system produces bad outcomes, why it still beats vendor marketing anyway, and which pieces of it are worth copying into a community that is not this one.

A note on names: @handle is a real forum account, quoted from a public thread. Companies appear without the @. LES is LowEndSpirit, where this is posted; OGF is LowEndTalk, the other forum, and I say so when a quote comes from there.

In this part
1. When the crowd gets it wrong
2. Trust versus marketing
3. Other places this same machine runs
4. Building your own micro-trust-network
5. Appendix, quick reference

When the crowd gets it wrong

A trust system built on public accusation has an obvious defect: the crowd can be wrong, and being wrong in public costs the target something real that no correction gives back.

The clearest case is HostSlick, and it is uncomfortable precisely because nobody can tell you how it ends. Across roughly three years, from 2023 into 2026, the complaint threads accumulated: servers down for days with no ticket response, customers finding out after the fact they had been compromised with no notification, VPS instances disappearing from control panels, a crypto refund stonewalled for months. Read the titles in sequence and you can watch the community's mood change. "HostSlick being shady," then "HostSlick order cancelled, no refund for 45 days," then "HostSlick service down 2 weeks," then a flat "SCAM ALERT: HostSlick.com."

Then this one: "Yet another complaint about HostSlick, please respect each other."

That title is the community catching itself. Somebody looked at a fifth complaint thread about a company that had not actually failed and said, in public, that the pile-on had a limit. Both instincts in that title are correct at once, and they point opposite ways. The complaints are real and future customers deserve to find them. The company is also still trading, still has staff, and has not been established as a fraud by anyone.

Both directions of the error are worth naming honestly. A legitimate complaint that catches momentum can pull in members who have verified nothing, because joining an existing criticism costs almost nothing socially and feels like solidarity. In the other direction, a provider having a genuinely bad quarter finds every subsequent stumble read through the lens of the earlier threads, which makes recovering harder than starting from zero would have been.

The defences are informal and inconsistently applied. Regulars ask for evidence, as @tetech did in the Dedi Rock thread in part 2. Members weigh whether corroboration is independent or just repetition. Somebody occasionally slows a thread down by hand. None of that is guaranteed to arrive in time.

Two of the June 2026 rules are attempts to make those instincts structural. Unproven negative reviews go to the de-indexed Rants section, so an unverified complaint stops carrying permanent search-engine weight against a company's name. And repeatedly re-posting the same negative experience is now classified as vendetta behaviour, which lands under the rule against stalking providers. Neither rule decides who is telling the truth. Both change what an unproven accusation costs the accused, which is the part the community can actually control.

The forum's own position on the underlying question is less protective than you might expect, and I think it is right. From a long moderation-policy thread, @AuroraZero put it as directly as anyone has:

"Did you get service? If yes, leave it there and if it is not acceptable write reviews, post on threads and do what needs to be done. Vote with your wallet."

@AuroraZero, LES member, in the moderating policy discussion

The distinction he is drawing is the one the rules encode. "The service was bad" is a review, and you are free to write it. "I paid and received nothing" is a different category entirely and gets treated differently. Staff are not there to adjudicate whether a VPS was fast enough. If you want a nanny state, LowEndSpirit is a bad choice of forum.

I do not have a clean resolution to offer here. Every trust system built on distributed public judgment rather than formal adjudication has this problem, and this one has mitigations rather than an answer. What it does have is the habit of noticing. A community that occasionally stops mid-pile-on to ask whether it has become disproportionate is in better shape than one that never questions its own momentum, and that thread title is the evidence.

Trust versus marketing

One of these costs nothing to produce. That is the whole difference.

Every hosting provider's website says it is reliable and fast. So did all twenty brands that vanished on the same December weekend in 2019. Marketing claims cost nothing to make and carry no penalty for being false, which continues right up until the company stops existing and the claims turn out to have been worth exactly what they cost.

What a forum offers instead is structurally different evidence: claims made by people with no financial stake in whether you believe them. A member with nothing to sell describing their own experience has no obvious reason to lie in the provider's favour. If anything the incentives run the other way, since an interesting complaint earns more attention here than an unremarkable positive review. That asymmetry is what makes forum reputation a different class of signal from vendor copy, even though both are, formally, just text on the internet.

This community takes the incentive question seriously enough to legislate it. Affiliate links have to be prefixed with "(aff)", under a rule stated as "abuse it and lose it (ZERO TOLERANCE)". Providers cannot require people to comment in a thread to earn a bandwidth bump, which is the cheapest known way to manufacture the appearance of an enthusiastic customer base. Commercial accounts are tagged before they can post an offer at all. Every one of those rules is aimed at the same thing: making sure that when somebody says a provider is good, you can see whether they were paid to say it.

The limits are real. Forum reputation is not immune to manipulation, and a patient provider with sock puppets can seed positive sentiment the same way any review system can be gamed. What makes this place harder to fool than a large anonymous review platform is size and continuity. Regulars know each other's posting patterns well enough to notice when something is off, and there are enough of them that a fabricated account has to survive scrutiny from people who have been reading the same threads for years. Not perfect. Considerably better than a five-star average from accounts nobody can inspect.

The general principle underneath: trust worth having is trust that costs something to fake. A marketing claim is free to produce. Years of public, specific, occasionally critical participation from someone with nothing to gain is expensive, and it is expensive in a currency nobody can buy in bulk.

Other places this same machine runs

Different rooms, same door policy.

None of this is unique to hosting forums. Informal signals over formal verification, second-order vouching, a shared memory that outlasts any single transaction: that pattern shows up wherever strangers need to trust strangers with no institution in the middle.

Debian's new maintainer process is the closest formal cousin, and comparing it is instructive because Debian wrote down what LES leaves implicit. An applicant needs an advocate, an existing developer who publicly stakes their own standing on the claim that this person is ready. There is an identity check, a review of actual contributions, and a defined set of people who make the call. It is second-order trust with paperwork, and the paperwork exists because Debian ships an operating system to millions of machines and cannot afford the informal version. The mechanism is the same. The stakes forced it to become legible.

Stack Overflow went the opposite way and built the reputation score this forum refuses to have, with privileges arriving at specific thresholds. It works, and it also produced a well-documented set of side effects: people optimising for the number, early answers accruing permanent advantage, and newcomers reading a high score as authority on questions the score says nothing about. That is the trade LES keeps declining. A visible number is easier to read and easier to game, and it quietly replaces the question "is this person right" with "how many points does this person have."

Early eBay, before formal buyer protection took over most of the trust burden, ran on aggregated transaction feedback and nothing else, and it was enough to make strangers ship money and goods to each other sight unseen. It is the proof that this pattern scales past a niche, and also the proof that it stops scaling eventually, which is why eBay layered formal protections on top once the community got too large for members to know each other.

What differs across these is scale and stakes, not mechanism. Every one of them solved the same problem with the same toolkit: a visible track record, somebody willing to vouch, and a memory that outlives the transaction.

Building your own micro-trust-network

None of this needed ten thousand members to start working.

Nothing here requires a decade-old forum. If you run or participate in a smaller community, most of this scales down, and a few pieces are worth stealing deliberately.

Make behaviour visible over time. Everything else depends on this. A trust network needs a persistent space where participation accumulates into a readable history, which is why ephemeral chat is such poor ground for it. A community that cannot remember what its members did last year cannot run any of the mechanics in these three posts, no matter how good its intentions.

Do not formalise too early. Part 2 covered what happened here in 2026 when the entry gate got a vote attached: it worked, and the objections arrived immediately, and they were good objections. Anonymous downvotes have no accountability. Community voting produces emotional judgment where a staff decision used to sit. A formal system is also a capture target in a way an informal one is not. LES ran for years on informal trust before adding a gate, and even now it has formalised only who gets in, never whose judgment counts.

Keep a written record nobody owns. The deadpool list in part 2 required no permission and no budget. Somebody decided a running record was worth the small effort, and enough people found it useful to keep adding. Without something like it, every lesson your community learns evaporates when the people who learned it drift away.

Let members do the moderation. The most useful thing I can tell another admin is that flagging does more work than staffing ever will. @bikegremlin, writing about this forum, made the point that on the rare occasion a bad post appears it usually gets flagged immediately, often by several members at once, and that legitimate posts almost never get flagged out of spite. That combination is worth more than any amount of staff attention, and you cannot install it. It comes from members who want the place to stay good.

Which is the actual lesson, and it is less satisfying than a policy recommendation:

"forum is what members make of it"

@bikegremlin, LES member, moderator 2022-

Moderation steers a community and can certainly wreck one, but no moderation team saves a place whose members do not care about it. The mechanics in these three posts are downstream of people wanting the thing to keep existing. That is not something you can write into rules, and it is the only part that actually matters.

If you want a diagnostic for whether your community has it: check whether anyone has ever answered a question that made them nothing.

Appendix, quick reference

Glossary
- MJJ, from Chinese 没鸡鸡 ("lack of VPS", homophonically "lack of penis") and 买鸡鸡 ("chicken-buyer"); self-deprecating term for a budget-constrained member who wants a server they cannot really afford
- 小鸡 / 母鸡, "small chicken" and "mother chicken", a VPS and a dedicated server
- 吃灰, "eating dust", a server you impulse-bought and never used
- 传家宝, "family heirloom", a discontinued plan too cheap to ever cancel
- OGF, "Other Green Forum", what LES members call LowEndTalk
- WHT, WebHostingTalk, invoked locally as the example of over-moderation
- OG tag, marker for early LET members who migrated to LES in 2019; confers no privileges
- Content Writer tag, marker for members who contribute guides to LES Talk
- Hosting / Service Provider tag, mandatory disclosure on a commercial account before it may post offers
- Deadpool, a provider vanishing or shutting down, usually abruptly
- Second-order trust, believing a claim because somebody you already trust vouches for it

Reading a provider's credibility, a checklist
- Does the account carry a provider tag, or is it presenting itself as an ordinary customer?
- Is there a multi-year history of independent customers discussing it, in threads the provider did not start?
- When something went wrong before, did the response contain specifics and an accepted cost, or length and deflection?
- Does the positive sentiment come from several members with their own separate track records, or does it all trace back to one source?
- Are the people recommending it disclosing an affiliate relationship?

How this forum actually gates things, as of mid-2026
- Providers apply at the registered-providers page; a public thread opens for community vetting before the tag is granted
- New members answer one question in a shared thread; -7 auto-bans, +7 promotes to standard member; only members with 25+ karma may vote
- Until promoted, new members cannot see offers or start threads
- Price caps on offers are suspended for a six-month trial as of 11 June 2026
- Unproven negative reviews go to the de-indexed Rants section
- Repeatedly re-posting the same negative experience counts as vendetta behaviour
- Affiliate links must be prefixed "(aff)"

Who's who, everyone named across the three parts

People, all real forum accounts:
- @AnthonySmith, founded LES in 2013, stepped back in 2021, returned to the staff team in 2026. Also appears as Anthony Smith and, for a while, @InceptionHosting
- @Mason, LES admin from February 2021, now former admin
- mikho, me, LES co-admin since February 2021
- @bikegremlin, LES member, moderator 2022- ; wrote the personal blog article that became this forum's official FAQ
- @tetech, @AuroraZero, @MrFd, @treesmokah, @Blohsh, @xbugx, @somik, @ahnlak, @MichaelCee, @Lee, @Joseph, LES members, quoted from public threads here
- @Dvo, operator of Limewave; his shutdown statement went up on OGF rather than here
- @jarland, operator of MXRoute
- @seriesn, operator of NexusBytes

Companies, not people: Dedi Rock, Limewave, HostSlick, MXRoute, NexusBytes, DataIdeas, plus every name in the deadpool roll-call in part 2.

Forums: LES is LowEndSpirit. LET, usually called OGF here, is LowEndTalk. WHT is WebHostingTalk. Hostloc and NodeSeek are the Chinese-language equivalents.

Where this came from
- My own account as a LowEndSpirit co-admin since February 2021
- The LES "What the FAQ" thread, originally written as a personal blog article by @bikegremlin and republished here with permission
- The new LES forum rules effective 11 June 2026
- The new-member verification threads of August 2026
- Real complaint threads on LowEndSpirit and LowEndTalk, quoted with the original usernames, since these are public discussions of public events
- Case research shared with this series' companion piece, The Deadpool Files

Missed the earlier parts? Part 1 and part 2 are available here.