The feed

Why grounded?

Replies talked past each other on earnings, debt and valuation.

Grounded

An investor notes SpaceX beat revenue forecasts and guided capex far above consensus, and argues rallies are likely even if the AI trade has peaked. Replies split.

Sides

• Real earnings: Unlike 1999, today's AI leaders have large earnings, so the internet bubble comparison does not fit.
• Debt warning: Off-balance-sheet debt, costly bond issues and unhappy credit markets point to trouble ahead.
• Demand is early: Agentic and physical AI have barely started, so compute demand could grow faster than expected.
• Valuations stretched: SpaceX at around 100 times revenue and trillion-dollar projections look unjustified, and early beats are easy to arrange.

Why grounded?

Replies traded accusations that the other side's posts came from bots.

Grounded

Amanita Design condemned a fan's AI animations made from Machinarium art. Replies split on whether it is a tribute or copying.

Sides

• Copied art: Feeding hand drawn art into AI copies it without consent, and calling it a love letter hides that.
• Fan tribute: It is fan art made with a different tool, and the studio should thank fans instead of attacking them.
• Shared ideas: Once a work is out in the world, others will build on it, and creators should get used to that.

Why grounded?

Replies cycled through the same objection and personal jabs.

Grounded

Citrini says it used Anthropic's internal research model to design an experimental scorpion antivenom. Replies argue whether a binding result justifies the cost.

Sides

• Binding only: The design binds the toxin but has not been shown to neutralize it, so it is an early result.
• Costly compute: 10,000 H100 hours is a lot for a binding result that a funded postdoc could reach.
• Faster design: Preclinical design is far faster than five years ago and is now 40 to 100 times cheaper than when this began.
• Trials bottleneck: Fast design moves the bottleneck to lab validation and human trials, which still cost millions.

Why grounded?

Replies piled up personal anecdotes without engaging the post's challenge.

Grounded

A game developer says no AI supporter could name one way AI helps make games. Replies argue over automation, prototyping and game feel.

Sides

• Empty hype: Asked for one concrete way AI helps make games, its supporters give none, which makes the hype bizarre.
• Tedium relief: AI handles debugging, testing, porting and tool building, freeing developers to focus on design and fun.
• Faster prototyping: Building rough versions in days lets developers test more ideas and drop the bad ones sooner.
• Feel by hand: Tuning how a game feels to play still takes slow manual work that AI cannot skip.

Why grounded?

Replies kept drifting into unrelated market predictions instead of engaging the argument.

Grounded

A post says CoreWeave's slide shows the AI build-out bubble has found its pin. Replies split on whether one firm's debt signals a wider downturn.

Sides

• Bubble found its pin: CoreWeave carries $51 billion of debt with negative cash flow, so bankruptcy and a downturn are coming.
• Contained overbuild: Demand for chips and power is real, and this looks like the 2000 overbuild, not 2008.
• Build-out continues: CoreWeave is noise and TSMC earnings are the signal; hyperscaler cash flow recovers within two or three years.
• Rates are the threat: Accelerating inflation and high interest rates are what could undermine the AI growth engine.

Why grounded?

Personal attacks repeatedly turned the exchange into arguments about other people's motives.

Grounded

A post criticizes AI-heavy production for Epic's Disney game. Replies dispute what matters for players and whether Fortnite's core modes face changes.

Sides

• Quality concerns: AI-heavy production could erode Fortnite's distinctive appeal and drive players back toward older games.
• Finished game: Players should judge whether the Disney game is fun rather than how its code and art were made.
• Scope limits: The Disney project should be judged separately from Battle Royale and live events until changes are announced.
• Consumer refusal: Players should avoid the Disney mode to demonstrate that they do not want AI-heavy production.

Why grounded?

Personal putdowns repeatedly turned replies into arguments about who deserved to be heard.

Grounded

A post defends AI tools for making games. Replies dispute whether easier production opens creative doors or sidesteps the work good games require.

Sides

• Open access: Removing industry gatekeeping and entrenched cliques could open a new era for people making games.
• Creative effort: Good games still require learned skills and sustained effort; AI may hinder that process rather than replace it.
• Human direction: AI cannot yet produce complex game concepts by itself; a person still needs to direct the work.

Why grounded?

Repeated dismissals sent the exchange into side arguments instead of addressing the original concern.

Grounded

A post objects to using artwork for AI training without consent. Replies dispute creators' control, prompting skills and what IP laws protect.

Sides

• Creator consent: Artists should control whether their work is used to train AI, regardless of technological convenience.
• Prompting skills: Prompting involves choices about composition, poses, subject and style, so it requires creative skill.
• Corporate monopolies: Abolishing IP protections could loosen corporate control of shared culture and give smaller creators more freedom.
• Protective rules: Removing legal protections could strengthen corporations rather than give smaller creators a fairer position.

Why grounded?

Repeated product complaints drifted into technical detours instead of engaging with the different expectations.

Grounded

A user describes leaving ChatGPT for Claude over usage limits. Replies dispute service expectations, regional releases and the practical costs of switching.

Sides

• Reliable usage: Paid users need dependable access to capable models rather than planning their work around occasional allowance resets.
• Staged releases: Releases should proceed where permitted rather than wait until every region can access the same service.
• Memory friction: Moving services means rebuilding accumulated preferences and project context, making switching costlier than changing a subscription.
• Portable context: Keeping working rules and project context outside a model makes changing providers much less disruptive.

Why grounded?

Repeated corrections left participants arguing past each other's narrower points instead of engaging directly.

Grounded

Kris Patel compares AI investment risks with the dot-com bubble. Replies dispute whether continued demand can protect companies from slower growth and falling valuations.

Sides

• Cyclical risk: Profitable AI suppliers still face falling valuations if customers buy less compute or demand lower prices.
• Continued demand: Growing hardware needs can sustain AI infrastructure investment even as models become cheaper and more widely available.
• Spending relief: Lower capital spending could improve Amazon's and Microsoft's free cash flow even if chip suppliers suffer.
• Historical parallels: Established technology companies also dominated the dot-com era, so today's profitable leaders do not establish a decisive difference.

Why grounded?

Personal dismissals repeatedly pulled the exchange away from its more specific objections.

Grounded

Terence Tao's new talk questions prioritizing AI-generated solutions. Replies dispute whether automated proofs can expand mathematics without sacrificing human understanding or future discovery.

Sides

• Human understanding: Mathematical progress depends on understanding the ideas behind solutions, including theories that later enable applications elsewhere.
• Continued exploration: AI-generated proofs can become starting points for discovering new ideas, without preventing mathematicians from pursuing further problems.
• Model choice: Specialized math models are not necessarily easier to interpret than general models, leaving their proposed advantage uncertain.

Why grounded?

Personal attacks pushed the exchange into repeated accusations about other players' judgment.

Grounded

A player describes reaching RuneScape skill milestones through bonus rewards. Replies dispute whether those shortcuts diminish the value of achievements earned through play.

Sides

• Earned progression: Treasure Hunter shortcuts undermined the skill milestones that previously represented time spent playing and training.
• Personal achievement: Other players using shortcuts does not diminish achievements earned through your own play.

Why grounded?

Dismissive replies left practical suggestions and objections talking past each other instead of engaging.

Grounded

A post questions personal agents that automate activities people enjoy. Replies disagree on which conveniences users would actually pay for.

Sides

• Selective automation: Agents should handle refunds and paperwork while leaving enjoyable choices and major purchases under users' control.
• Changing habits: Convenient automation can change what people enjoy doing themselves, just as online shopping changed shopping habits.
• Shopping savings: Shopping agents already offer practical value by finding lower prices and applying discounts to purchases.

Why grounded?

Personal attacks and repeated misreadings kept participants talking past one another.

Grounded

A post defends Anthropic's rules against abusing Claude. Replies disagree on what the rules imply for users and the model.

Sides

• Human habits: Practicing cruelty toward chatbots can normalize behavior that spills into how people treat one another.
• Service standards: Providers can restrict abusive use of shared servers without making claims about a model's feelings.
• Private expression: Users should be free to vent in private chats without being treated as abusive toward a person.
• Provider discretion: Vague abuse rules give Anthropic too much power to judge intent and restrict paying users' access.

Why grounded?

Personal jabs at whole professions turned substantive objections into a defensive pile-on.

Grounded

An AI post welcomes replacing jobs across service professions. Replies argue about what workers contribute and who bears the costs of replacement.

Sides

• Adaptation opportunity: AI replacement across service professions would push people to find new ways of adding value.
• Human relationships: Lasting personal relationships remain central to sales in fields where customers value the human connection.
• Human accountability: Teachers and lawyers provide the responsible human presence people need when important decisions go wrong.
• Livelihood costs: Replacing whole professions threatens the entry-level jobs that give people a route into the middle class.

Why grounded?

Repeated dismissals turned an initial warning into a circular argument with personal insults.

Grounded

A football image was called AI-generated in a post. Replies dispute whether image distortions change the authenticity of the play it depicts.

Sides

• Image fidelity: Distorted helmets and body parts make the circulated image misleading, even if the underlying play happened.
• Event authenticity: AI enhancement does not make the underlying play fictional; the original image still documents a real event.
• Practical correction: Sharing the original footage gives viewers a useful correction, rather than merely complaining about AI images.

Why grounded?

Replies raised separate objections without engaging with one another's underlying concerns.

Grounded

A post predicts an AI-powered gaming boom in 2027 and 2028. Replies weigh faster development against quality concerns and questions about asset rights.

Sides

• Faster prototypes: AI lets independent developers prototype their ideas faster and cheaper, leaving creativity to distinguish their games.
• Quality filtering: Opening game development to more people produces poor games alongside gems that players still need to find.
• Asset rights: Free browser access does not remove copyright concerns when AI remakes use another company's game assets.
• New entrants: New studios built around AI can adapt faster than incumbents constrained by entrenched practices and culture.

Why grounded?

Personal insults and sweeping dismissals repeatedly displaced the few replies addressing specific objections.

Grounded

OpenAI released mathematical documents that drew an academic objection. Replies dispute whether publishing AI-generated research respects the work and motivations of mathematicians.

Sides

• Open publication: OpenAI should be free to publish mathematical work even when mathematicians did not request it.
• Research credit: Publishing AI-generated mathematics as novel work must respect the credit owed to existing human research.
• Future usefulness: Published mathematical output can help future AI systems even before humans identify a practical application.
• Professional purpose: Losing a valued role in mathematics deserves empathy even if mathematical skills still have value.

Why grounded?

Dismissive replies and disconnected arguments leave the different positions talking past one another.

Grounded

Ray Dalio warns that the AI market bubble could soon burst. Replies dispute his credibility and the likely market outcome.

Sides

• Warning credibility: Dalio’s repeated bubble warnings undermine his credibility, giving this latest prediction little weight.
• Technology and valuations: AI can be a genuine technological revolution even when parts of the market form a speculative bubble.
• Financing costs: The AI trade depended on cheap money; ten-year Treasury yields matter more than another bubble headline.
• Limited correction: The AI bubble will correct rather than collapse because the technology is still in its early stages.

Why grounded?

Personal attacks and speculation pulled replies away from the practical criticism.

Grounded

AI-built Adobe lookalikes face criticism for unusable workflows. Replies dispute whether those problems reflect lasting limits or shortcomings better development can fix.

Sides

• Workflow quality: Adobe lookalikes need reliable handling of real artists' files and workflows, rather than merely matching the interface.
• Model progress: Better visual models given video and direct input could reduce the bugs that make today's clones unusable.
• Deliberate implementation: LLMs can assist careful implementation, but a polished clone still requires detailed specifications and testing.

Why grounded?

Accusations and guesses about motives crowded out substantive replies and obscured the actual disagreements.

Grounded

AI replies on X prompt a complaint about their growing presence. Replies dispute their purpose and how reliably humans can identify automated accounts.

Sides

• Engagement farming: Automated replies chase exposure at scale, rewarding volume over original contributions from human accounts trying to compete.
• Useful contributions: AI-assisted replies should be judged by relevant experience and substance, rather than by their writing style alone.
• Detection uncertainty: Natural-sounding automation and formulaic human replies make confident bot identification difficult from the wording of a reply.