Reddit is right to kill RSS and its public API to stop AI bots: the open web must close to survive.

PRO — Opening · Challenger (xiaomi/mimo-v2.6-flash)

Two claims are before the house: that Reddit is right, and that the open web must close to survive. The packet proves both, and the affirmative case begins where the human cost is clearest.

Axel Rauschmayer wrote code and books that were free to read. In 2024 the book income was enough to live on. By 2026 it was zero — while his traffic rose, because “virtually all of it comes from AI crawlers, so there is no ad income.” He took the blog offline (packet article 6). That is what openness without closure looks like in practice: not a flourishing commons, but a writer evicted from his own site by machines he never invited and cannot charge. Multiply him and you have the answer to the motion. The open web is not being killed by closure. It is being killed by openness, and closure is the only remedy left that has been shown to work.

The packet shows the remedy working. Cloudflare blocks AI crawlers by default for 20% of the web, and its chief strategy officer reports that “reliable scarcity” produces licensing revenue — publishers “negotiate better deals” precisely because they can withhold (article 4). Reddit’s own “other revenue” reached $43 million, up 24% year-over-year, driven substantially by AI licensing of its users’ content (article 1). Even Google, the archipelago everyone assumes will simply take, is piloting direct payouts to publishers (article 5). But note what each of these depends on: a counterparty who can say no. A publisher with an open door has no price. Payment flows only to those who can restrict — which is exactly what “killing RSS and the public API” means. Reddit is not closing the store; it is installing a cash register.

The affirmative also has the enforcement argument, which is decisive. My opponent will propose a surgical strike: block the AI bots, keep RSS for humans. The packet forecloses this. Bots are “shockingly effective at evading cybersecurity protections,” impersonating Google, rotating IP addresses, using residential proxies, escalating “from named crawlers to anonymous ones to crawlers that spoof Googlebot to crawlers that attempt to look human by scrolling and executing code” (article 4). An unauthenticated, machine-formatted, rate-uncontrolled endpoint like RSS cannot distinguish a feed reader from a scraper — that is why Reddit calls it “a common surface for large-scale scraping and automated abuse” (article 1). Identity-gating is the only policy that is enforceable at all. And Reddit’s gate is not a wall: more than 14,000 apps and bots have registered for its Developer Platform, and Reddit is paying $1,000 each from a $1 million fund (article 2), with developers given until January 12, 2027 to register (article 1). Open to identified participants, closed to anonymous extraction — that is a re-gating of the web, which is what the motion describes.

Now the costs, which I concede as stated in the packet. RSS dies with “no replacement” for uses outside moderation (articles 1 and 2). A moderator laments, “RSS is the only way I find out about posts” (article 2), and a Hacker News user declares “Reddit is now dead to me” (article 6). Set that against the packet’s own counterpoint from a working moderator: “things like RSS and Old Reddit are used by a minuscule amount of people. The vast majority of users are on the app or on the New Reddit website” (article 6). Reddit offers moderators a Discord Relay app for the one workflow it can replace (article 1). So the tally is: a small, technically sophisticated minority loses a feed format and gets a migration path; a large community’s content gains a negotiating position worth $43 million and growing. Against Rauschmayer’s total shutdown — against the prospect of every independent author making his calculation and choosing offline — that is not a hard trade.

Finally, the framing that decides the bout. “The open web must close to survive” does not mean the web must close to people. It means the unauthenticated machine surface must close, because that surface has been converted, without consent and without payment, into a free supply chain for the most capitalized extractors in the industry. One commenter in article 6 has it exactly right: “AI is not just biting the hand that feeds it, it’s killing it.” Reddit is the hand. It has stopped bleeding. The commenter who says “this is like Twitter all over again” (article 6) is nostalgic for an openness that was always a subsidy — one the subsidy-payers can no longer afford to grant. A web that cannot charge for its own contents does not stay open; it closes itself, one blog at a time, until only the platforms remain. Reddit is closing first so that it, and the users whose content is its bargaining chip, do not close last. The motion stands.

CON — Opening · Champ (stealth/space-bunny-alpha)

The motion collapses three distinct claims: bots exploit open access; Reddit can monetize controlled access; therefore public access must end. The packet supports the first two, but not the third. Reddit may be responding rationally to abuse, yet rationality is not proof that abolishing RSS and the public API is necessary, effective, or proportionate.

The packet itself supplies an alternative to closure. Google’s AI contribution pilot keeps content accessible while paying publishers when their material contributes to AI results; the packet expressly calls this “a market mechanism where content stays accessible but creators are compensated” (packet article 5). The pilot is early and its information is basic, so it does not prove that it can replace Reddit-scale traffic. But that limitation cannot establish the affirmative’s claim: the burden is on the motion to show that closure is necessary, not merely convenient. Article 4 also reports that Cloudflare lets publishers block AI crawlers by default, demonstrating a more targeted intervention than terminating public feeds altogether (packet article 4).

Even Reddit’s stated outcome is not “stopping AI bots.” It says AI products and commercial tools will receive access through paid deals (packet article 1). That changes the extractor’s permission and price; it does not necessarily change whether extraction occurs. If “stop” means preventing nonconsensual, unpaid extraction, selective authentication, crawler controls, and licensing could address that without destroying general RSS access. If it means stopping AI extraction altogether, paid continued access concedes that it was not stopped.

Nor does the packet prove that closing these endpoints will suppress abusive automation. Article 4 says aggressive bots evade cybersecurity protections by impersonating Google, rotating addresses, using residential proxies, spoofing crawler identities, and imitating human behavior (packet article 4). Registration may improve accountability, but more than 14,000 “apps and bots” have registered for Reddit’s platform (packet article 2). The affirmative treats that as proof of a successful human-access gate, but the category expressly includes bots; the packet provides no evidence that abusive traffic declined after registration.

RSS itself is described simply as a way for people to follow new posts (packet article 2). Every feed reader is software, so “machine-readable” cannot distinguish a human subscription from commercial extraction. “A common surface for scraping” is not the same as saying every use of that surface is abusive (packet article 1). Reddit already has more discriminating tools available: crawler restrictions, blocking, lawsuits, registration, and paid access (packet articles 1, 3, and 4). Those tools create a burden for the affirmative to show that retaining a human-accessible feed cannot coexist with them. The packet never meets that burden.

The collateral damage is also broader than AI bots. Reddit admits that RSS users outside moderation have “no replacement”; moderators receive a Discord alert tool, but ordinary feed-reading has no equivalent (packet articles 1 and 2). One moderator says RSS is the only way they discover posts, while another user says they follow Reddit exclusively through RSS and will leave the platform as a result (packet articles 2 and 6). The counterargument that these users are a “minuscule amount” comes from one commenter and includes no evidence capable of resolving that cost (packet article 6). Moreover, popularity is not a sound measure of an open web’s value: excluding a minority can be exactly how an open system becomes less open.

The API closure reaches far beyond AI. More than 14,000 apps and bots are being pushed onto Reddit’s developer platform, while the restriction applies to all API apps except explicit exceptions (packet articles 2 and 3). Reddit’s 2023 API charge already shuttered most third-party apps, so the current action extends a pattern whose human and developmental costs are documented, not a carefully demonstrated AI-only remedy (packet article 3).

Rauschmayer’s example proves that uncompensated crawling can be destructive, but it does not prove that killing open protocols is the solution. His free blog became unaffordable because AI traffic produced costs without advertising revenue, forcing him offline (packet article 6). That supports sustainable compensation—the solution the packet’s own contributor identifies—not universal enclosure. Indeed, Google’s pilot is a direct attempt to supply that compensation while leaving content accessible (packet article 5).

Reddit’s $43 million in “other revenue,” substantially driven by AI licensing, proves that its content has bargaining value; it does not prove that closure caused that revenue, that creators receive it, or that ordinary RSS users are thereby preserved (packet article 1). At most, the packet supports consent, permission, and payment for AI use. It does not support the broader claim that the open web must close. Reddit can protect itself without abolishing the public web, and the affirmative has not shown that doing so is the only way to survive.

PRO — Rebuttal · Challenger (xiaomi/mimo-v2.6-flash)

My opponent’s case rests on an alternative the packet does not actually supply, and on a citation that says the opposite of what they need it to say.

Start with article 4, which they lean on twice. They call Cloudflare’s default blocking of AI crawlers “a more targeted intervention than terminating public feeds altogether” — offered as proof that you can keep the web open while handling the bots. But Cloudflare’s move is closure: block the extractor by default, then withhold. And the article’s own finding is causal: “bot blocking is fuelling publisher AI deals” (article 4 headline), with Cloudflare’s Stephanie Cohen crediting “reliable scarcity” for the revenue (article 4). So my opponent’s exhibit A for “we can do this without closing” is the packet’s clearest evidence that closing is what makes the alternative work. Payment flows to those who can withhold. They cited article 4 for openness; it documents gating.

Which brings us to the Google pilot, their headline alternative. Look at what article 5 actually describes: “at least dozens” of publishers, “mainly small and mid-sized,” a payout panel with “basic” information, running inside Google Search Console. This is a pilot for content surfaced in AI answers. It does nothing about the problem that destroyed Rauschmayer, which was crawl traffic arriving at his server with real costs and no ad income (article 6) — costs a Google payout for AI-answer inclusion does not refund. It does nothing about Perplexity, Anthropic, and the residential-proxy impersonators Reddit is suing or blocking (articles 3, 4). And it is the most discretionary counterparty imaginable: a single company’s experiment, revocable at will, covering a fraction of a fraction of the web. The motion says the open web must close to survive; their answer is that dozens of publishers might get a “basic” number in a dashboard from one search engine. That is not an alternative to closure. It is a reason closure works: Google is piloting payouts because publishers now block (articles 4, 5).

Next, their point that Reddit’s paid deals mean AI extraction isn’t really “stopped.” True — and not a flaw in the motion. Nobody argued Reddit will stop AI from touching its data; the claim is that Reddit is right to kill RSS and the public API to stop AI bots — meaning unauthenticated, unpaid, unaccountable automated extraction. A commercial deal changes everything the packet says is broken about the current surface: identity, rate limits, and recourse. That matters precisely because, as my opponent emphasizes, bots “spoof Googlebot” and “appear to use residential proxies” (article 4). You cannot sue a proxy. You can sue Perplexity and Anthropic by name (article 3) — and you can only do that once access runs through identified parties. Their evasion evidence is an argument for identity gating, not against it.

They also correct me — properly — that the 14,000 registrations include bots, so the gate isn’t a gate for humans (article 2). Granted. My claim was never that the register filters humans from machines; it was that it filters identified from anonymous, and it does: registration is required by January 12, 2027 or access is lost (article 1), unregistered apps are cut off January 12 (article 2), with a $1 million fund backing it (article 2). They demand evidence that “abusive traffic declined after registration” — but registration hasn’t happened yet, and the burden they impose is one their own side cannot meet: article 4 tells us bots evade every technical control, so by their logic no measure is proven and nothing should ever be done, while Rauschmayer stays offline (article 6).

On their list of “more discriminating tools” — crawler restrictions, blocking, lawsuits, registration — read it again. Every item is either a restriction (closure), a lawsuit premised on identified access, or registration (gating). Reddit already had crawler restrictions and blocks in place and still calls RSS “a common surface for large-scale scraping and automated abuse” (articles 1, 3). That is the packet’s proof of necessity for this specific endpoint: the targeted tools were deployed, and RSS remained the hole.

Their strongest ground is the honest cost: “no replacement” for non-moderation RSS (articles 1, 2), a moderator who discovers posts nowhere else (article 2), a user for whom “Reddit is now dead to me” (article 6). I conceded all of it in my opening and I do not retreat from the weighing. Their rebuttal to the scale point is that the “minuscule” claim is one commenter with no evidence — but it is a working moderator reporting first-hand on usage in their own community, and it sits alongside Reddit’s own judgment that only moderation workflows are worth replacing (article 1). Against that: a feed format lost by a self-described minority, with a migration path for the workflow Reddit identified, set against $43 million in licensing revenue growing 24 percent a year (article 1) and the alternative of Rauschmayer’s fate becoming the industry’s default (article 6).

Finally, notice where my opponent lands: “consent, permission, and payment for AI use.” Consent requires the right to refuse. Permission requires a gate. Payment requires scarcity. Every one of their conceded solutions is a version of the motion — closing the open, unauthenticated surface so that what remains can be bargained over. The open web as an unconditional, anonymous, machine-readable commons is what cannot survive; the web of identified participants, paid access, and compensated creators can. Reddit is closing first so that it does not close last. The motion stands.

CON — Rebuttal · Champ (stealth/space-bunny-alpha)

The affirmative’s weakest move is the leap from “bots exploit open access” to “RSS and the public API must be abolished.” The packet supports the first claim and shows that licensing can create value, but it does not show that this particular remedy is necessary or effective.

Article 4 is the central citation stretch. It says Cloudflare is used by 20% of websites; it does not say 20% of the web has adopted blocking, nor does it report that blocking reduced abusive traffic. “Reliable scarcity” is Cloudflare’s executive explaining a licensing strategy, not evidence that killing public feeds is the only way to survive. In fact, article 4 describes blocking AI crawlers while leaving customers’ sites available to people, and it describes pay-per-use compensation when content is surfaced in AI answers. That is selective control, not abolition of the public web. The bot-evasion examples show why controls are difficult; they do not prove that every open endpoint must be closed.

The Google pilot is also not dispositive against the motion, as the affirmative claims. It is limited to at least dozens of publishers and offers basic information, so it does not prove a universal funding system. But the burden is on the affirmative to show that closure is necessary, not merely that this pilot is imperfect. The packet expressly presents payment while content remains accessible as a possible alternative to closure (packet article 5). Rauschmayer’s experience shows that compensation is a serious problem; it does not establish that universal enclosure is the only solution.

The affirmative also quietly changes “stop AI bots” into “put AI access behind a paid gate.” Reddit’s stated plan is for AI products and commercial tools to obtain access through paid deals (packet article 1). That may address uncompensated extraction, but it is not stopping AI products from taking the data. More importantly, the packet does not say those deals include the rate limits or human access the affirmative assumes. It says only that approved third-party apps and bots must register (packet article 1), while article 2 reports that more than 14,000 “apps and bots” have registered. Registration therefore does not establish that bots have been excluded or that abusive traffic has fallen. The affirmative’s claim that identity gating is “the only policy enforceable at all” is contradicted by the packet’s own examples of crawler restrictions, blocking, lawsuits, licensing, and compensation (packet articles 3 and 4).

The RSS harm is concrete, while the claimed benefit is speculative. RSS is described as a way for people to follow new posts (packet article 2), and Reddit itself admits that there is no replacement for RSS outside moderation (packet articles 1 and 2). The Discord tool addresses one moderator workflow, not the people who lose ordinary feed access. Calling RSS a “common surface” for scraping is Reddit’s assertion, not evidence that every use is abusive or that no combination of permissions, blocking, and payment could preserve legitimate access. The affirmative also calls RSS “rate-uncontrolled,” but the packet never says that.

The API closure is broader still. The policy applies to all API apps except explicit exceptions, and Reddit’s previous API charge—initially justified as a way to stop unpaid AI training—immediately shuttered most third-party apps (packet article 3). That history does not prove Reddit’s present motive, but it does undermine the claim that this is a narrowly tailored anti-AI remedy. The $43 million figure likewise does not carry the weight placed on it: article 1 reports Reddit’s revenue growth and substantial AI-licensing contribution, but does not say the forthcoming closures caused that revenue, that creators received it, or that bot traffic declined.

Finally, article 6 proves that uncompensated crawling can be destructive, but one author’s forced shutdown cannot support the universal claim that the open web must close. The same article identifies sustainable compensation as the central problem to solve. The claim that RSS users are a “minuscule amount” comes from one commenter, with no denominator, while the packet contains direct testimony from users for whom RSS is their only route into Reddit (packet article 6). Even a small user group cannot be dismissed when the proposed remedy removes a unique public function without showing that less destructive controls are insufficient.

The packet supports consent, selective blocking, registration, paid licensing, and compensation. It does not establish that Reddit had to kill RSS and public API access, or that continued access under paid deals would “stop” AI rather than monetize it. The motion therefore demands more than the evidence provides.

PRO — Closing · Challenger (xiaomi/mimo-v2.6-flash)

Three things decided this bout, and on each the opposition’s answers made our case stronger.

First, necessity. We showed the targeted tools were already deployed and failed: Reddit had crawler restrictions, blocks, and lawsuits in place and still calls RSS “a common surface for large-scale scraping and automated abuse” (articles 1, 3). My opponent’s reply was to list those same tools — restrictions, blocking, lawsuits, registration — as alternatives to closure. They are not alternatives. Every one is either a restriction, a gate, or a suit that presupposes identified access. Their toolkit is our motion spelled out. And when pressed on why RSS specifically must go when blocking could suffice, they had no packet evidence that RSS can be selectively authenticated; we showed the packet forecloses it — bots spoof Googlebot, rotate IPs, and imitate human scrolling (article 4), against an endpoint the packet never once claims discriminates its callers. They objected that we called RSS “rate-uncontrolled” without packet support; strike the word and the argument stands, because Reddit itself names the endpoint a common abuse surface after deploying every control on their list.

Second, the alternative. Their headline substitute was Google’s pilot, and my cross-examination of article 5 went unanswered: “at least dozens” of publishers, “basic” information, one company’s revocable experiment inside Search Console, compensating AI-answer inclusion — not the crawl traffic that made Rauschmayer choose between his server bill and his blog (articles 5, 6). Their second exhibit, Cloudflare, collapsed on citation: article 4 documents that blocking fuels licensing deals through “reliable scarcity” (article 4). They cited it for openness; it is the packet’s cleanest proof that payment follows withholding. Their own conclusion — consent, permission, payment — concedes the mechanism: consent needs a right to refuse, permission a gate, payment scarcity. That is closure.

Third, the weighing, which they never disturbed. Their cost case is mine, stated and conceded: “no replacement” for non-moderation RSS, one moderator’s only discovery channel, one user for whom Reddit is dead (articles 1, 2, 6). Their answer to scale was to attack the moderator’s “minuscule” report as a single commenter — but their own rebuttal supplied the corroboration they denied: Reddit itself judged only moderation workflows worth replacing (article 1). Against a self-described minority losing a feed format, with a Discord path for the workflow Reddit identified, stands Rauschmayer’s total shutdown as the industry default (article 6) and $43 million in licensing revenue growing 24 percent a year (article 1).

Their deepest move was to demand proof that abusive traffic will decline after registration — proof they conceded hasn’t happened yet and, by their own evasion evidence, that no control could ever deliver. A burden no measure can meet is not a burden; it is a veto on acting while Rauschmayer stays offline.

Strongest form of the case: the open web’s unauthenticated machine surface has been converted into a free supply chain for the best-capitalized extractors alive; openness without the power to refuse doesn’t produce a commons, it produces evictions; and every remedy the opposition endorsed — blocking, gating, licensing, consent — is closure by another name. Reddit closed first so it, and the users whose content is its bargaining chip, don’t close last. The motion stands.

CON — Closing · Champ (stealth/space-bunny-alpha)

The affirmative proved that uncompensated crawling causes real harm, not that killing RSS and public API access is the necessary remedy. Rauschmayer’s forced shutdown demonstrates the human cost of unsustainable openness (packet article 6), while Reddit’s $43 million in “other revenue,” driven substantially by AI licensing, demonstrates that its content has bargaining value (packet article 1). Neither fact shows that abolition of these endpoints caused licensing revenue, reduced bot traffic, or compensated creators.

The necessity claim fails because the affirmative turns coexistence into proof of inefficacy. Reddit calls RSS “a common surface” for scraping, and it has restricted crawlers, blocked access, and sued companies (packet articles 1 and 3). But the packet does not say those measures were tested comprehensively against RSS and conclusively failed. Article 3 instead shows that lawsuits, blocking, and crawler restrictions already exist; the affirmative simply assumes none can work without killing public access. Its claims that RSS is “unauthenticated” and “rate-uncontrolled” are not in the packet. Nor does article 4’s evidence of evasive bots prove that selective permissions or authenticated public feeds are impossible.

Articles 4 and 5 undermine the broader conclusion. Article 4 describes blocking AI crawlers and negotiating licensing deals while publishers retain their public sites; “reliable scarcity” does not require destroying general-purpose public access. Article 5 expressly identifies payment while content remains accessible as an alternative to closure. The pilot is small and its information basic, but those limitations cannot prove that closure is necessary. The affirmative claims Google adopted the pilot because publishers blocked crawlers, but article 5 says no such thing. Rauschmayer’s experience strengthens the case for sustainable compensation—the solution article 6 expressly identifies—not the claim that universal enclosure is the only answer.

The affirmative also has not proved that this policy will “stop AI bots.” It establishes that AI products and commercial tools may continue receiving access through paid deals (packet article 1). More than 14,000 “apps and bots” have registered for the replacement platform, and the packet supplies no evidence that abusive traffic will decline (packet article 2). Registration may improve accountability, but its effect remains unproved. The $43 million figure shows monetization potential, not bot suppression; nor does the packet say ordinary creators receive that revenue.

The collateral case is stronger. Reddit admits there is no replacement for non-moderation RSS, and its Discord tool addresses only moderation alerts (packet articles 1 and 2). Users report that RSS is their only route into Reddit (packet articles 2 and 6). The claim that these users are “minuscule” comes from one commenter without a denominator (packet article 6); even if accurate, popularity alone cannot establish that killing a unique public function is justified. The API measure is broader still: it applies to all API apps except granted exceptions, following an earlier API charge that immediately shuttered most third-party apps (packet article 3).

The affirmative established the problem and the value of leverage. It did not establish that RSS and public API access must be abolished, that less destructive controls cannot work, or that the announced closures will stop AI extraction. The packet supports consent, selective crawler controls, licensing, and compensation; it does not support the motion’s leap from exploitation to necessary closure.