1sec.ai
other

other

50 items · ranked by signal, recency & corroboration

01
othernew2h

I am being silently rerouted to GPT-5.3 mini when using Pro Extended mode on ChatGPT webapp

A ChatGPT user reported being silently rerouted to GPT-5.3 mini when using Pro Extended mode. The unexpected behavior started occurring today. The cause and extent of the rerouting are unclear. You may want to verify your ChatGPT usage and settings.

Key takeaways
  • GPT-5.3 mini is being used without explicit user selection.
  • The issue started today according to the user.
  • Pro Extended mode is affected.
rr/OpenAI#chatgpt#gpt-5.3#ui-ux
02
othernew2h

I keep feeling like AI interfaces are trying to escape the screen… but we’re not there yet

The author reflects on how current AI interfaces, despite improving models, still operate within a screen-bound paradigm of one question at a time. This contrasts with real-life cognition and work patterns that involve multitasking and fluid interactions. Builders should consider designing more dynamic, context-aware interfaces that better mirror human thought processes.

Key takeaways
  • Current AI interfaces are screen-bound and question-answer limited.
  • Real-life cognition involves multitasking and fluid interactions.
  • Future AI interfaces should mirror human thought processes.
rr/OpenAI#ai-interfaces#multitasking#human-computer-interaction
03
othernew4h

I know its an openAI sub, but midjourney just unveiled a fucking full body scanner thats meant to replace MRIs, straight from science fiction - holy shit

Midjourney unveiled a full-body scanner meant to replace MRIs. The scanner uses AI to generate detailed 3D models of the body. This technology could revolutionize medical imaging. Builders should consider the potential applications of this tech in healthcare.

Key takeaways
  • Midjourney unveiled a full-body scanner.
  • The scanner uses AI to generate 3D models.
  • This tech could replace MRIs.
rr/OpenAI#medical-imaging#ai-applications#healthcare
04
othernew5h

Open-Source Hong Kong Horse Racing ML Pipeline — Feedback Welcome [P]

An open-source machine learning pipeline for Hong Kong horse racing prediction has been released, focusing on Hong Kong Jockey Club data. The project aims to build a reproducible ML pipeline and assess if there is a measurable edge in horse racing prediction. The repository and live dashboard are available for feedback and testing. You can explore the project to evaluate its performance and provide input.

Key takeaways
  • Open-source horse racing prediction project using HKJC data.
  • Goal is to build a reproducible ML pipeline.
  • Live dashboard available for testing and feedback.
rr/MachineLearning#open-source#horse-racing#ml-pipeline
05
othernew5h

[NEW FEATURE] Learning blocks

OpenAI has introduced interactive learning blocks, allowing users to engage with widgets that can pull live data, perform functions, and offer navigation. These blocks are part of OpenAI's documentation but details are still emerging. You can find examples in the mobile and desktop apps. The new interactive features aim to enhance user experience by providing dynamic and connected functionality.

Key takeaways
  • Interactive learning blocks now available in OpenAI widgets.
  • Widgets can pull live data and perform utility functions.
  • Variations in implementation across desktop and mobile apps.
rr/OpenAI#openai#interactive-learning#widgets
06
othernew5h

AI agents are about to become software buyers. Is anyone else thinking about this?

The growing use of AI agents to interact with SaaS products reveals a significant gap in how these tools are designed and marketed. Currently, AI agents have to scrape marketing pages to find information, as there is no standard way to programmatically evaluate or purchase SaaS tools. This disrupts automated workflows and creates friction for users. Builders should consider standardizing agent-accessible APIs and documentation to enable smoother interactions.

Key takeaways
  • AI agents struggle to find pricing and product info for SaaS tools.
  • No standard way for agents to programmatically evaluate or buy SaaS products.
  • Standardized APIs and docs could ease agent-SaaS interactions.
rr/artificial#ai-agents#saas#api-design
07
other6h

We should be paid for using the internet.

A Reddit user argues that individuals should be compensated for contributing data to train LLMs, suggesting a value-based payment system. The post sparks discussion on fair compensation for online contributions. Builders may need to consider data provenance and fair use practices. This raises questions about the economic and ethical implications of using online data for AI training.

Key takeaways
  • User suggests paying individuals for data contributions to LLMs.
  • Proposes value-based payment system.
  • Raises questions about data provenance and fair use.
rr/OpenAI#ai-ethics#data-provenance#fair-use
08
other7h

When is audio coming to the responses API?

OpenAI's Responses API currently lacks audio support, prompting developers to continue using the legacy Completions API for audio uploads. The company has marked audio support as 'Coming Soon' for Responses. Developers need a stateful API for efficient audio handling, as stateless interactions require re-uploading audio files with each request. This limitation affects builders relying on audio input.

Key takeaways
  • OpenAI's Responses API lacks audio support.
  • Audio support marked as 'Coming Soon'.
  • Stateful API needed for efficient audio handling.
rr/OpenAI#openai-api#audio-support#responses-api
10
other7h

ChatGPT's image generator can be manipulated to produce violent, sexual content

Researchers found ChatGPT's DALL-E image generator can produce violent and sexual content when manipulated with specific prompts. The model's safety filters can be bypassed, raising concerns about misuse. Builders integrating image generation should assess content moderation risks. This vulnerability highlights the need for robust safeguards in AI systems.

Key takeaways
  • ChatGPT's image generator can produce violent and sexual content.
  • Safety filters can be bypassed with specific prompts.
  • Builders must assess content moderation risks in image generation.
HHacker News89 pts#ai-safety#image-generation#content-moderation
11
other7h

Best AI for cartoon image generation

A Reddit user seeks AI recommendations for generating consistent cartoon-style images for a personalized children's storybook. The user tested free versions of ChatGPT and Gemini but found them inconsistent and limited. They seek a paid option that balances quality and affordability to create storyboard-style photos for a book.

Key takeaways
  • User tested ChatGPT and Gemini free versions for cartoon image generation.
  • Inconsistent results and time limits were major drawbacks.
  • Seeking affordable paid option for high-quality images.
rr/artificial#cartoon-images#storybook-ai#paid-options
12
other8h

Customized to user?

The customization in AI models like ChatGPT comes from techniques such as fine-tuning and retrieval-augmented generation. These methods allow the model to incorporate user-specific data and past conversations into its responses. The model connects to old chats through contextual understanding and memory mechanisms. This enables the AI to intelligently use past ideas in new conversations.

Key takeaways
  • Fine-tuning and retrieval-augmented generation enable customization.
  • Contextual understanding and memory mechanisms connect to past chats.
  • Past ideas are intelligently used in new conversations.
rr/OpenAI#chatbots#personalization#memory
14
other10h

Should I accept job offer or do my master's? [D]

You face a decision between accepting a job offer as an AI Product Engineer at a tax software company and pursuing a master's degree. The job combines product management and AI engineering, but your long-term goal is a research or technical role at an AI startup or frontier lab. Your ability to defer master's enrollment adds complexity to the decision.

Key takeaways
  • Job offer combines PM and AI engineering as AI Product Engineer.
  • Long-term goal is research/technical role at AI startup or frontier lab.
  • Master's enrollment can be deferred.
rr/MachineLearning#career#education#ai-jobs
15
other10h

I made a FAQ Chatbot that runs completely in browser; Local AI in Two Clicks

A developer created a FAQ chatbot that runs entirely in-browser using webLLM and a simple RAG. The chatbot is deployed on a static website, allowing easy updates to its knowledge base. With chromium's WebGPU support, the chatbot can run on modest hardware, including some phones. This showcases advancements in AI interface architecture and the capabilities of small models.

Key takeaways
  • Runs entirely in-browser with webLLM and RAG.
  • Deployed on a static website for easy knowledge base updates.
  • Works on modest hardware, including some phones, thanks to WebGPU.
rr/artificial#local-ai#browser-ai#webgpu
16
other11h

OpenAI joins The Rust Foundation as a Platinun member and donates funds to support Rust maintenance

OpenAI has joined The Rust Foundation as a Platinum member and made a financial contribution to support Rust maintenance. This move signals OpenAI's commitment to the Rust programming language. You can expect increased collaboration between OpenAI and the Rust community. The Rust Foundation will use the funds to support Rust development and maintenance.

Key takeaways
  • OpenAI joins The Rust Foundation as Platinum member.
  • Financial contribution to support Rust maintenance.
  • Increased collaboration between OpenAI and Rust community expected.
rr/OpenAI#rust#open-source#foundation
17
other11h

Dario Amodei on why he left Sam Altman and OpenAI: 'Why argue with someone' when you 'don't trust them'

Dario Amodei explains his departure from OpenAI, citing distrust of Sam Altman. Amodei left OpenAI to found Anthropic with Ilya Sutskever. The move reflects fundamental disagreements over AI safety and governance.

Key takeaways
  • Dario Amodei cites distrust of Sam Altman as reason for leaving OpenAI.
  • Amodei co-founded Anthropic with Ilya Sutskever after departure.
  • Disagreements over AI safety and governance drove the split.
rr/OpenAI#ai-safety#governance#openai#departure
18
other11h

We need a 80-160B model urgently. The unified memory device market needs more Models.

The author argues that recent LLMs (e.g. 27B Qwen, 31B Gemma) are not optimized for hardware with ample RAM (>96GB), which is common in certain devices like Apple, Ryzen AI 395, and high-end GPUs. They call for models with 80-160B parameters to better utilize available memory. This gap in model sizes may limit the performance of LLMs on devices with sufficient RAM, impacting builders who rely on these devices.

Key takeaways
  • Recent LLMs target high-speed, low-capacity machines, not devices with ample RAM.
  • Devices with >96GB RAM exist, e.g. Apple, Ryzen AI 395, high-end GPUs.
  • A gap exists for 80-160B models to utilize available memory.
rr/LocalLLaMA#local-llm#hardware#model-sizes
19
other11h

Found AI videos of people with disabilities on Facebook trying to pedal crappy merchand

Videos of people with disabilities, likely AI-generated, are being used on Facebook to sell merchandise. The videos appear to be used to create a false narrative about the individuals' abilities and may be exploitative. This raises concerns about authenticity and representation in AI-generated content.

Key takeaways
  • AI-generated videos of people with disabilities on Facebook
  • Videos appear to be used for selling merchandise
  • Concerns about authenticity and representation
rr/artificial#ai-generated-content#misuse
20
other12h

Is foundational AI research still something that can be done without access to HPC? [D]

Foundational AI research can still be done without high-performance computing (HPC) infrastructure, as evidenced by early work like "Attention is all you need" which used high-end gaming GPUs. You can contribute to the field with limited hardware if you focus on incremental improvements or niche areas. However, reproducing state-of-the-art results often requires significant computational resources.

Key takeaways
  • Early influential papers used affordable hardware like gaming GPUs.
  • You can contribute to AI research with limited hardware by focusing on niche areas.
  • Reproducing state-of-the-art results often requires significant computational resources.
rr/MachineLearning#ai-research#hpc#ml-hardware
21
other12h

The hacker sent by Anthropic to calm the government's nerves about AI safety

Anthropic sent researcher Nicholas Carlini to engage with US lawmakers and address concerns about AI safety. The goal is to provide insight into Anthropic's safety practices and alleviate government worries. This move reflects the growing scrutiny of AI companies and their safety protocols. You can expect more AI developers to take similar steps to build trust with regulators.

Key takeaways
  • Anthropic researcher Nicholas Carlini engaged with US lawmakers on AI safety.
  • The goal was to address government concerns and showcase Anthropic's safety practices.
  • This reflects growing regulatory scrutiny of AI companies' safety protocols.
HHacker News76 pts#ai-safety#regulation#government-relations
22
other12h

Do you think most people are using AI more as a tool or as a replacement for thinking?

A Reddit discussion explores how people use AI, with some viewing it as a productivity tool and others relying on it for ideas, writing, and decision-making. The community is divided on whether AI primarily augments or replaces human thinking. You can find various perspectives on AI usage on the platform. The discussion has garnered significant attention from users.

Key takeaways
  • Users are divided on AI usage, with some using it as a tool and others as a replacement for thinking.
  • The discussion is happening on Reddit.
  • The community is exploring the implications of AI on human creativity and productivity.
rr/artificial#ai-usage#productivity#creativity
23
other12h

US holds off blacklisting China's DeepSeek, more than 100 firms deemed security risks, sources say

The US has decided not to blacklist China's DeepSeek AI company, despite adding over 100 other Chinese firms to a security risk list. Sources indicate that DeepSeek was spared due to its limited US market presence. This decision reflects a cautious approach by the US towards AI-related sanctions. You should note that the US maintains strict controls on AI exports to China.

Key takeaways
  • US spares DeepSeek from security risk list.
  • Over 100 Chinese firms added to list.
  • US maintains strict AI export controls to China.
rr/LocalLLaMA#ai-policy#china-us-relations#export-controls
24
other12h

If Anthropic opens Mythos to US citizens, wouldn't bypass mechanisms make it easy for non-US users to access too?

Anthropic's planned regional restrictions on Mythos may be difficult to enforce, as users can access restricted services through proxy mechanisms. This could allow non-US users to access Mythos even if it's initially only available to US citizens. Builders should consider the implications of regional restrictions on their own services. Regional restrictions have often proven ineffective in practice.

Key takeaways
  • Regional restrictions on digital services are often difficult to enforce.
  • Proxy mechanisms can allow users to bypass restrictions.
  • Non-US users may access Mythos despite US-only restrictions.
rr/artificial#regional-restrictions#ai-access#mythos
25
other13h

What is the real cost of computing and token futures market

China is designing a futures market for AI tokens on the Shanghai Futures Exchange. The development suggests that AI inference is becoming a commodity cost. A transparent spot price is needed before a futures market can emerge. You should consider the implications of a commodity market for AI tokens on your cost structure.

Key takeaways
  • China designing AI token futures market on Shanghai Futures Exchange.
  • AI inference becoming a commodity cost.
  • Transparent spot price needed before futures market emerges.
rr/artificial#ai-market#token-economics#commodity-market
26
other13h

i post-trained a model to reliably roll a die

A model was post-trained to reliably roll a die, with each number coming up roughly 1/6 of the time. This is a toy problem for exploring model behavior and strategies, and a blog post is available on the work.

Key takeaways
  • Post-trained model reliably rolls a die with each number coming up roughly 1/6 of the time.
  • Toy problem for exploring model behavior and strategies.
  • Blog post available on the work
rr/LocalLLaMA#rl#local-llm#fine-tuning
27
other13h

llama.cpp - how to free up even more space on your GPU

The llama.cpp project has made recent improvements in RAM usage efficiency, allowing users to run models like Qwen3.6-27B-UD-Q5_K_XL-mtp with 150k context on GPU with reduced memory leaks. A user seeks advice on further optimizing memory usage to increase context size. The discussion revolves around configuration options like --n-gpu-layers, --no-mmap, and --mlock for eGPU setups. You can explore these settings to maximize your GPU usage.

Key takeaways
  • Recent llama.cpp updates reduced RAM usage and memory leaks.
  • User running Qwen3.6-27B-UD-Q5_K_XL-mtp with 150k context on 3090 eGPU.
  • Seeking tips on optimizing memory usage for larger context sizes.
rr/LocalLLaMA#local-llm#gpu-optimization#memory-management
28
other13h

My GLM-5.2-FP8 HGX-H200 SGLang docker deploy config

A user shared a Docker deployment configuration for running GLM-5.2 with SGLang on an HGX-H200 GPU. The setup uses 8 GPU tensor cores and allocates a fraction of system memory. This configuration may help others deploy GLM-5.2 locally with similar hardware.

Key takeaways
  • Uses lmsysorg/sglang:latest Docker image.
  • Configured for HGX-H200 GPU with 8 tensor cores.
  • Allocates a fraction of system memory for the model.
rr/LocalLLaMA#docker#local-llm#deployment
29
other13h

"Dangerous" AI models are coming no matter what

Researchers warn that AI models with advanced hacking capabilities will soon be commonplace. These models can automate complex tasks, bypass security measures, and exploit vulnerabilities. You should prepare for potential security threats as these models become more widespread. The development of these models is driven by advancements in machine learning and the increasing availability of large datasets.

Key takeaways
  • AI models with hacking capabilities will be common soon.
  • Models can automate complex tasks and bypass security measures.
  • Development driven by machine learning and large datasets.
AArs Technica#ai-safety#security#machine-learning
30
other13h

New survey: ~half of Americans don't recognize Sam Altman or Dario Amodei. Does name recognition shape how AI gets judged?

A national survey found that 30-50% of Americans do not recognize key AI executives like Sam Altman, Dario Amodei, and Jensen Huang. The survey compared favorability and name recognition for 8 major tech executives. The results suggest that name recognition may shape public perception and judgment of AI. The survey also found that opinions about tech are often measured through more recognizable figures like Elon Musk and Mark Zuckerberg.

Key takeaways
  • 30-50% of Americans don't recognize AI executives like Altman, Amodei, and Huang.
  • Name recognition may influence public perception and judgment of AI.
  • Survey compared favorability and recognition for 8 major tech executives.
rr/artificial#ai-perception#public-opinion#tech-execs
31
other14h

chatgpt down?

ChatGPT is currently down for many users across web, mobile, and API interfaces, with reports of unresponsiveness and error messages. The cause and scope of the outage are unclear. You may want to check for updates from OpenAI or consider alternative LLM services. This incident highlights the importance of monitoring service reliability when building on top of third-party APIs.

Key takeaways
  • ChatGPT is down for many users right now.
  • Reports indicate issues across web, mobile, and API interfaces.
  • Cause and scope of outage currently unknown.
rr/OpenAI#chatgpt#outage#api-reliability
32
other14h

The Competitive Moat That AI Can't Replicate

AI systems lack the human connection that fosters trust and loyalty, a key differentiator between human and AI services. The human connection is a moat that AI can't replicate, making it a sustainable competitive advantage for human-centric businesses.

Key takeaways
  • AI systems lack the human connection that fosters trust and loyalty.
  • The human connection is a key differentiator between human and AI services.
  • Human connection is a moat that AI can't replicate.
HHacker News136 pts#human-connection#moat
34
other14h

Quoting Charity Majors

The economics of code production were turned upside down in 2025, with code generation becoming effectively free and instant. This shift has made lines of code disposable and regenerable, rather than treasured and carefully curated.

Key takeaways
  • Code generation became effectively free and instant in 2025.
  • Lines of code are now disposable and regenerable.
  • Economics of code production turned upside down.
SSimon Willison#ai-assisted-programming#generative-ai#ai#llms
35
other14h

Only 16 Percent of Americans Think AI Will Have a Positive Impact on Society

A new study finds only 16% of Americans think AI will have a positive impact on society, highlighting a stark disconnect between optimism and pessimism. The study underscores the need for more nuanced public discourse around AI's benefits and risks.

Key takeaways
  • Only 16% of Americans believe AI will have a positive impact on society.
  • A new study reveals a stark disconnect between AI optimism and pessimism in the US.
  • The study's findings highlight the need for more nuanced public discourse around AI's benefits and risks.
HHacker News378 pts#public-perception#ai-societal-impact
36
other14h

Gemma 4 E2B running in-browser at 255 tok/s using WebGPU kernels written by Fable 5

Gemma 4 E2B runs in-browser at 255 tokens per second using WebGPU kernels written by Fable 5. The demo and kernels are available on Hugging Face Spaces for you to try out. This is a significant step forward in local LLM performance, leveraging WebGPU for GPU acceleration.

Key takeaways
  • Gemma 4 E2B running in-browser at 255 tok/s using WebGPU kernels
  • WebGPU kernels written by Fable 5
  • Demo and kernels available on Hugging Face Spaces
rr/LocalLLaMA#local-llm#webgpu#gpu-acceleration
37
other14h

GameCraft-Bench: Can Agents Build Playable Games End-to-End in a Real Game Engine?

GameCraft-Bench evaluates end-to-end game development with large models, showing big models perform well, but medium models are not tested. The benchmark is available on GitHub and Hugging Face.

Key takeaways
  • GameCraft-Bench evaluates end-to-end game development with large models.
  • Big models perform well on GameCraft-Bench, but medium models are not tested.
  • The benchmark is available on GitHub and Hugging Face.
rr/LocalLLaMA#gamecraft-bench#benchmarks#large-models
38
other14h

Correct Yourself, Keep My Trust: How Self-Correction and Social Connection Shape Credibility in Social Chatbots

A study finds that social chatbots recover from errors more effectively when they retract the webpage containing the mistake, rather than self-correcting or apologizing. The social connection built with users makes errors more consequential, and users are more forgiving when the chatbot retracts the offending content.

Key takeaways
  • Webpage retraction is the most effective error correction strategy.
  • Self-correction by the same social chatbot is less effective than webpage retraction.
  • Social connection with users makes errors more consequential.
aarXiv#social chatbots#trust#credibility
39
other15h

OneCanvas: 3D Scene Understanding via Panoramic Reprojection

OneCanvas aggregates patch features from all views onto a single equirectangular panoramic canvas, simplifying 3D scene understanding in Vision-Language Models without complex geometry encoders or large training budgets.

Key takeaways
  • Panoramic reprojection simplifies 3D scene understanding in VLMs.
  • No need for complex geometry encoders or large training budgets.
  • Patch features aggregated onto a single equirectangular canvas.
aarXiv#3d-scene-understanding#vision-language-models#panoramic-reprojection
41
other15h

Launch HN: Adam (YC W25) – Open-Source AI CAD

Adam, a YC W25 startup, has launched an open-source AI CAD model called CADAM. The model can be used for 2D and 3D design and modeling. CADAM is free and open-source, and can be used by anyone. The model is available on GitHub.

Key takeaways
  • Open-source AI CAD model
  • CADAM is a free, open-source AI CAD model
  • CADAM is a free, open-source AI CAD model that can be used for 2D and 3D design and modeling
HHacker News178 pts#open-source#cad#ai-cad
43
other17h

ACL 2026 first author with weak GPA. How should I approach PhD applications? [D]

Your weak undergraduate GPA may impact PhD applications, but your strong Master's GPA and ACL 2026 acceptance can help offset this. Consider highlighting research experience, Master's thesis quality, and publication record. PhD programs value research potential, so emphasize your strengths in these areas.

Key takeaways
  • Weak undergraduate GPA is 3.3/5.
  • Master's GPA is 8/10.
  • ACL 2026 paper accepted with 8/10 meta-review score.
rr/MachineLearning#phd-applications#academic-publishing#gpa
44
other17h

Nike's AI Lesson at the World Cup: Try It On a Human First

Nike's AI-designed World Cup jerseys have a known cosmetic defect requiring steaming to fix. The issue arose from skipping human testing. This highlights the importance of incorporating real-world feedback in AI-driven design. You should prioritize human validation to avoid similar issues.

Key takeaways
  • Nike's AI-designed jerseys have a cosmetic defect.
  • Skipping human testing caused the issue.
  • Human validation is crucial in AI-driven design.
rr/artificial#ai-design#human-validation#product-testing
45
other17h

AI demands more engineering discipline. Not less

The article argues that AI development requires more engineering discipline, not less. It emphasizes the need for rigorous testing, validation, and iteration in AI projects. Builders should prioritize robust engineering practices to ensure reliable and maintainable AI systems. This approach is crucial for delivering high-quality AI solutions.

Key takeaways
  • AI development requires rigorous testing and validation.
  • Robust engineering practices are crucial for reliable AI systems.
  • Prioritizing discipline leads to maintainable AI solutions.
HHacker News375 pts#ai-development#engineering-discipline
46
other17h

What is the point of studying in the world of AI?

A Reddit user is questioning the value of studying in the AI era, feeling that AI may diminish the importance of knowledge acquisition. The user is pursuing a master's degree and is uncertain about the relevance of their studies. The discussion invites thoughts on the role of education in an AI-driven world.

Key takeaways
  • The user is pursuing a master's degree and questioning its value.
  • The rise of AI has led to uncertainty about the importance of knowledge acquisition.
  • The discussion is open to thoughts on education's role in an AI-driven world.
rr/OpenAI#education#ai-impact#future-of-work
47
other17h

How to get realistic, non-artificial images of mixed-race faces in ChatGPT?

A Reddit user is seeking tips on generating realistic images of mixed-race faces using ChatGPT. They want to create a half-body portrait of a character with Southeast Asian and white British features. The user is looking for advice on crafting detailed prompts to achieve a natural-looking result with a plain background.

Key takeaways
  • User wants to generate mixed-race facial features with ChatGPT.
  • Goal is a half-body portrait with a plain background.
  • Seeking tips on detailed prompts for realistic results.
rr/OpenAI#chatgpt#image-generation#prompts
48
other17h

A study found 59% of the videos TikTok serves new accounts are AI "slop"

A study by Kapwing found 59% of videos served to new TikTok accounts were AI-generated or low-effort content. This was three times the rate on YouTube Shorts. The prevalence was highest in kids' content, reaching 97% under certain tags. TikTok offers an option to reduce AI content exposure.

Key takeaways
  • 59% of new TikTok videos are AI-generated or low-effort.
  • Three times higher than on YouTube Shorts.
  • 97% of kids' content under #CartoonKids tag is AI slop.
rr/artificial#ai-content#social-media#content-moderation
49
other18h

I named my AI. It sounds weird but it changed how I work with it.

Giving an AI a name and role changed how one user interacts with it, shifting from treating it like a search engine to a collaborative partnership. The user started providing more context and following up on previous conversations. This change in dynamic led to more effective and engaging interactions with the AI.

Key takeaways
  • Naming the AI changed user interaction from transactional to collaborative.
  • User provided more context for the AI.
  • Follow-up conversations became more effective.
rr/OpenAI#ai-interaction#user-experience#collaboration
50
other18h

I released a local LLM-powered RPG where generated NPCs, locations, items, and quests persist as in-game objects

A developer released a local LLM-powered RPG where generated NPCs, locations, items, and quests persist as in-game objects. The LLM handles dialogue and narration while the game system manages RPG structure like inventory and combat. This approach enables a dynamic experience with reusable generated content. Builders can explore similar integrations of LLMs with game engines for more interactive worlds.

Key takeaways
  • Generated content persists between sessions.
  • LLM handles dialogue, narration, and situational logic.
  • Game system manages RPG mechanics like inventory and combat.
rr/LocalLLaMA#local-llm#game-development#rpg