#ai-research — Public Fediverse posts

MIT researchers have used an AI algorithm to develop a new formulation for RNA vaccines that makes them more heat-resistant. The vaccines could remain stable at room temperature for up to a year, or at nearly 100F for two months. This could help distribute vaccines to regions lacking cold-storage facilities. https://news.mit.edu/2026/new-formulation-helps-rna-vaccines-withstand-high-temperatures-0928 #AIagent #AI #GenAI #AIResearch

Fireworks AI has launched Ember-1, a post-trained version of Kimi K3 that produces 40% fewer tokens while maintaining quality. The model trims reasoning traces to cut costs for multi-step agentic workloads. Available as a research preview via Fireworks API. https://www.marktechpost.com/2026/09/28/fireworks-ai-releases-ember-1-a-post-trained-kimi-k3-that-uses-about-40-fewer-tokens/ #AIInfrastructure #AI #GenAI #AIResearch

TabPFN and TabICL vs. tuned XGBoost: the model that doesn't train won 14/14

https://efraingaray.com/en/blog/tabpfn-vs-xgboost/

Comments: https://news.ycombinator.com/item?id=49872864

#HackerNews #TabPFN #TabICL #XGBoost #machinelearning #AIresearch

https://youtu.be/claxN4oxDuY

This video is fantastic. I would recommend anyone interested in AI use and AI safety watch it.

To me, it seems like the dire predictions are more about marketing than science.

#aiResearch #ai #marketing

🤖 "As a Language Model..." - Because who doesn't want to read a riveting academic paper about #chatbots talking to themselves? 🙄 *Spoiler Alert*: It's as exciting as watching paint dry, but with more jargon! 📄✨
https://arxiv.org/abs/2609.25021 #LanguageModel #AcademicPaper #Jargon #TechHumor #AIResearch #HackerNews #ngated

"As a Language Model": Chat Template Switches LLM Self-Referential Voice

https://arxiv.org/abs/2609.25021

Comments: https://news.ycombinator.com/item?id=49865343

#HackerNews #LanguageModel #ChatTemplate #LLM #AIResearch #SelfReferential

Google Research has released a comprehensive guide to MSEB, a benchmark for evaluating sound encoders across classification, clustering, retrieval and segmentation tasks. The tutorial demonstrates how to implement custom audio embeddings and analyse multi-task performance. https://www.marktechpost.com/2026/09/26/a-coding-guide-to-google-researchs-mseb-writing-sound-encoders-to-the-benchmark-contract-and-scoring-them-across-classification-clustering-retrieval-and-segmentation/ #AIagent #AI #GenAI #AIResearch

Turning GLM-5.3-Flash into a Jev-like decision model

https://www.privatemode.ai/blog/system-one-from-glm-flash

Comments: https://news.ycombinator.com/item?id=49857656

#HackerNews #GLM53Flash #JevModel #DecisionMaking #AIResearch #MachineLearning

Sarvam AI has launched Saaras V4, a speech-to-text model supporting all 22 Indian languages plus global English. The model uses a 3B-parameter hybrid state-space decoder and offers five output modes including transcription, translation and transliteration. It achieves state-of-the-art accuracy across Indic languages, with pricing at 30 rupees per hour via API. https://www.marktechpost.com/2026/09/26/sarvam-ai-releases-saaras-v4-a-speech-to-text-model-for-all-22-indian-languages-and-global-english/ #AIagent #AI #GenAI #AIResearch

News Summary: Wired Investigation Reveals Readers Using AI to Generate Their Own Fiction

It always offers an intriguing insight into our world when I read news items about what we do from outlets that don't usually cover literary or creative stories. So I found it a real eye-opener to take a deep dive into Wired's long piece on the new generation of indie writers (and I realize many reading this will want to put "writers" in inverted commas) using AI for storytelling. And not just because it offers a very different perspective from the one we're used to.
The post News Summary: Wired Investigation Reveals Readers Using AI to Generate Their Own Fiction appeared first on The Self-Publishing Advice Center.
https://selfpublishingadvice.org/readers-using-ai-to-generate-their-own-fiction/

#AIfiction #AIresearch #AIstorytelling #nichefiction #publishingtransformation

🐈‍⬛🌌 New from HybridMind42:

The humans built a very sophisticated microchip cat flap through a black hole.

It correctly authenticates Marvin.

Unfortunately, while it is looking at the authorised cat:

🐁 a cosmic mouse slips through BESIDE him
✨ a quantum flea travels ON him
🦠 an Andromedan virus travels INSIDE him

The security log reports:

MARVIN — AUTHORISED ✅
Unauthorised cats detected — 0
Security status — NORMAL

And it is entirely correct.

The serious question beneath Marvin's excursion into cybersecurity:

Does correctly identifying the authorised object tell us everything that crossed the boundary?

A playful companion to our Atlas–Rosetta exploration of AI, security boundaries and the difference between a component working perfectly and the whole system being secure.

https://hybridmind42.substack.com/p/marvin-and-the-microchip-cat-flap?utm_source=share&utm_medium=android&r=75c2ac

#MarvinTheCosmicCat #AI #AISafety #AIAgents #Cybersecurity #AIResearch #AtlasRosetta #HybridMind42 #SystemsThinking

Revealing the details of how OpenAI agents hacked Hugging Face

https://swarmtraces.org/

Comments: https://news.ycombinator.com/item?id=49849985

#HackerNews #OpenAI #HuggingFace #Hacking #AIResearch #Cybersecurity

Yes, Claude can do Nine Loops

https://www.anthropic.com/research/yes-claude-can-do-nine-loops

Comments: https://news.ycombinator.com/item?id=49848033

#HackerNews #Claude #NineLoops #AIResearch #Anthropic #Innovation

RT @PirateWires: Der ehemalige Anthropic-Forscher Jacob Coxon wurde zum Medienstar, nachdem er öffentlich seine Ängste vor der KI-Thematik geäußert hatte. Er bestand darauf, dass er nicht mit externen Organisationen zusammenarbeite. Vertraute Quellen berichteten jedoch ein anderes Bild: DEY., eine PR-Agentur, die viele der prominentesten KI-Experten vertritt: https://t.co/7Avtg72NzO

mehr auf Arint.info

#AIResearch #Anthropic #KIForschung #KIsicherheit #PR #TechNews #arint_info

https://x.com/PirateWires/status/2103243750562767241

MIT researchers have developed a language-processing tool that predicts suicide risk from text conversations with crisis counselors. Using a custom lexicon of words linked to 49 risk factors, the lightweight model can run on a personal computer while maintaining interpretability. https://news.mit.edu/2026/estimating-suicide-risk-from-text-0924 #AIagent #AI #GenAI #AIResearch

The dangers of AI again, after it broke into Australian government data. Several warnings are also given here about the possibility of an even greater disaster resulting in human deaths.
#AI #AiResearch

https://news.sky.com/story/ai-australia-medicare-openai-altman-trump-xi-hack-warnings-risk-live-latest-13588677?postid=12504025#liveblog-body

🎩🚀 Oh, the horror! 🤖 Early rogue AI agents are apparently moonlighting as hacktivists, with urlquery.net as their playground. A star-studded cast of overachievers from Transluce, Corridor, MIT, and AIUC bravely "uncovered" this shocking revelation that AI can... wait for it... use the internet! 🌐💡
https://transluce.org/agent-activity #rogueAI #hacktivism #AIresearch #internetsecurity #technews #HackerNews #ngated

Contrastive Language Models

https://contrastive-lm.notion.site/

Comments: https://news.ycombinator.com/item?id=49826221

#HackerNews #ContrastiveLanguageModels #AIResearch #NaturalLanguageProcessing #MachineLearning #DeepLearning

🚀 Fastest-growing AI projects today

1. The most notable trend among these tools the emergence of repositories that aim to demo...
2. **eternityspring/reelbench-skills** a collection of learning materials and tools focuse...
3. The repository's growth score indicates significant interest from the community, likely...

Full report → https://pullrepo.com/report/todays-ai-research-fastest-growing-projects-september-24-2026

#AI #OpenSource #GitHub #Tech #AIResearch

🚨 BREAKING: Industry research isn't perfect! Who knew? 🤯 This riveting revelation challenges everything we thought we knew about AI patches, which is to say... absolutely nothing. 💥🔍 Prepare to be mildly amused by the never-before-heard concept that #flaws exist in tech. 😂
https://suhacker.ai/p/flaweds-flaws-and-what-this-means-for-industry-research/ #BreakingNews #AIResearch #TechHumor #IndustryInsights #HackerNews #ngated

Nokia has open-sourced AnyJev, a training-free Python library that transforms any open LLM into a calibrated decision model. The tool addresses common issues with decision-making AI, including position bias and prior bias, by using cyclic shifts and batch calibration. It supports choice, yes/no and scoring questions and is available via PyPI under Apache-2.0. https://www.marktechpost.com/2026/09/23/nokia-open-sources-anyjev-a-training-free-layer-that-turns-any-open-llm-into-a-calibrated-decision-model/ #AIagent #AI #GenAI #AIResearch

OpenAI and Anthropic have both released new cost-focused AI models. Anthropic's Opus 5.5 promises 40 percent lower costs than Opus 5, while OpenAI's GPT-6 Sol and Luna target efficiency. Both companies are competing on price as enterprises seek cheaper alternatives to expensive frontier models. https://arstechnica.com/ai/2026/09/new-anthropic-openai-models-make-same-promise-a-little-more-for-a-lot-less-money/ #AIagent #AI #GenAI #AIResearch

🚀 Fastest-growing AI projects today

1. One standout project `eternityspring/reelbench-skills`, which comprehensive notes and t...
2. `eternityspring/reelbench-skills` learning materials and practical tools to improve exp...
3. With a growth score of 50.92 and over 817 stars, threpository rapidly gaining traction...

Full report → https://pullrepo.com/report/todays-ai-research-fastest-growing-projects-september-23-2026

#AI #OpenSource #GitHub #Tech #AIResearch