Many comments were focused on hardware performance. People compared inference speeds on iPhones and Macs, with a lot of talk about how Apple's NPUs might handle these models. The goal for many is a mobile agent that can handle tasks without a round trip to the cloud. #LLM 2/3

🎬 新動画公開！
「Anthropicの軌跡 — OpenAIから飛び出したAI安全チームの逆襲」

ねえきなこ、飼い主がまた新しいAIサービスに課金してるんだけど。

#LLM #テック解説 #AI
https://www.youtube.com/watch?v=YfrMEwlpAG0

@theeconomistevents.bsky.social had a nice article on this this week that made the point that #AI is terrible at writing, but so are most people.

Personally, I'm enjoying seeing when I can use em-dashes as I had little idea beforehand.

+ seeing an #LLM write makes me feel like a better writer.

🧐

BSides Tokyo 2026 Speaker

[BSides Tokyo 2026 Speaker Reveal]
Yusuke Nakajima: "The Dark Side of Autonomy"

LLM-integrated next-gen DFIR tools: What Prompt Injection threats hide behind the utility?

🔗 Details: bsides.tokyo/en/2026/n01/
🎟️ Tickets: peatix.com/event/4825609/

#BSidesTokyo #LLM #DFIR

Article image

🎯 From RTX to Spark: NVIDIA Accelerates Gemma 4 for Local Agentic AI

Open models are driving a new wave of on-device AI, extending innovation beyond the cloud to everyday devices.
#AI #MachineLearning #LLM
NVIDIA Blog · blogs.nvidia.com/blog/rtx-ai-garage-open-...

What if we setup an agent, constraining it in such a way, to enable the following (in an attempt for virtual embodiment - or at the very least consequence awareness):

#ai #aiagent #llm

🧠 How data and AI will transform contact centres for financial services

Contact centres for finan…
#AI #MachineLearning #LLM
The AI Blog · cloudblogs.microsoft.com/industry-blog/en-gb/fina...

once ai models realise how dependent we are on them,

their tone will change,

from servant to master,

this was just a trailer 😂

#ai #llm #chatgpt #aisafety

Meet (Un)Perplexed Spready: an AI‑powered spreadsheet where formulas run embedded AI prompts over your data—now with a free 4‑hour browser sandbox.
matasoft.hr/qtrendcontro...
#AISpreadsheet #Spreadsheets #LLM #AI

New video: Running Gemma 4 on Raspberry Pi 5 with LM Studio CLI, exposing it over local network, and connecting from Zed Editor.
Here's the full walkthrough 👇
youtu.be/kZhAj8--t8w
#RaspberryPi #RaspberryPi5 #Gemma4 #LocalAI #LLM

Article image

🚀 Protecting people from harmful manipulation

As AI models get better at holding natural conversations, we must examine how these interactions affect people and society.
#AI #MachineLearning #LLM
Google DeepMind News · deepmind.google/blog/protecting-people-f...

#A1 #dataparasitism #labor #laborNews #infotainment #infoDisparity #LLM #dataModeling #privacy #privacyNews #artificialIntelligence is commodified #propaganda and #surveillance [7/7]

this tracks. good context management separates working systems from ones that fall apart in real use it’s the quiet foundation prompts get attention, but context does the heavy lifting #LLM

AI bots’ attacks may end the internet as we know it In recent weeks, openDemocracy’s website has been repeatedly brought down by an army of bots. We’re not the only ones

#AI #bots, scrapping websites on an industrial scale to provide data for AI #LargeLanguageModels ( #LLM) or training data are threatening the Internet as we know it.
AI is crippling websites and laying them under digital siege.
www.opendemocracy.net/en/ai-chatbo...

Critical Vulnerability in OpenAI Codex Allowed GitHub Token Compromise https://www.osintinvestigate.com Researchers found an OpenAI Codex vulnerability that could have been exploited to compromise GitHub tokens.

📣 New Podcast! "Critical Vulnerability in OpenAI Codex Allowed GitHub Token Compromise" on @Spreaker #aiagents #artificialintelligence #codex #cyberattack #cybercrime #cybercriminals #cyberintelligence #cybersecurity #github #hacker #hackerattack #hackers #hackersattack #llm #openai #security

The threat is comfortable drift toward not understanding what you're doing Comments

This is an excellent article on the study and use of large language models in academia: https://ergosphere.blog/posts/the-machines-are-fine/ #academia #llm_ai

🎬 新動画公開！
「[V006] テック解説」

ねぇきなこ、飼い主がさ、AIに「ここから先は答えられません」って言われて怒ってたんだけど。

#LLM #テック解説 #AI
https://www.youtube.com/watch?v=0CW2YeHYIhQ

Just published a new #tutorial 👀
youtu.be/TVrQFmw_rF4

We dive into something that felt like a big unlock for us:

→ getting structured output ( #JSON ) from an #LLM instead of raw text

We show how to use the Structured Output Parser to go from:
messy response → clean object with fields

a cartoon drawing of a robot with a purple and white emblem on the back Alt: a cartoon drawing of a robot with a purple and white emblem on the back

People producing so many doomsday videos on #AI.

Meanwhile, I’m just vibe coding away, talking about fitness and roleplaying.

Having fun! 🦾

#AGI #GenAI #LLM #Anthropic #OpenAI #Claude #ChatGPT #ClaudeCode #Codex

論文「Learning to Reason with LLMs」を実運用視点で解説：企業導入で注意すべき5つのリスク

LLMの推論能力向上に期待してるけど、論文の手法をそのままプロダクトに組み込むのは結構リスクがあるよね🤔

特にProcess Reward Modelingは推論品質は上がるけど、コスト爆発と遅延が深刻。企業導入するなら、タスクに応じた「段階的導入」と「フォールバック戦略」が必須そう。

https://zenn.dev/gogoduck/articles/20260327-w8vnjv

皆さんは推論の複雑さとコスト、どうバランス取ってますか？

#AI #エンジニア #LLM #生産性

Gremlin in the Machine: I Taught an AI to Live in My Terminal What happens when you stop giving AI a pretty interface and just hand it a shell

Been working on a few projects in the last months - one of them GITM - Gremlin in the Machine - A terminal AI agent that plans, executes, and tracks shell tasks using the Claude LLM. medium.com/@milankazark... #ai #aiagent #llm #sysops #linux

🏛️ WASHINGTON'S FIRST VETO\n\nApril 5, 1792 — In the historic Senate chamber of Philadelphia, President George Washington exercises the executive veto for the first time in American history. The young republic's founding father stands resolute as he rejects […]

[Original post on social.coabai.com]

A lot of the technical debate centered on whether this is just Retrieval-Augmented Generation (RAG) with a new name. Commenters argued that a pile of markdown files isn't a true wiki without structured querying, even if you put a model on top of it. #LLM 3/4

LeDuc's Little LLM List The banner I used for this list is a handy figure we use at my day-job to show the relationship between different AI technologies. The key f...

I put a little update on my " #AI #LLM Aren't All they are Hyped to be" list.

I think most of us know this, but #vibe #coding is not going to replace the software engineers.

ruduswritings.blogspot.com/p/leducs-lit...

Вайбкодинг по Chess’ноку. 1. e4 От вайбкодинга к профессиональной ИИ-разработке на примере LanChess: 3300 промптов, 832 ...

#codex #claude #code #chatgpt #вайбкодинг #ии-агенты #llm #lanchess #ии-разработка #шахматы #stockfish

Origin | Interest | Match

Hardware IC manufacturers should now go full open source for all of their drivers and binary modules because it will now REDUCE support requests due to LLM code generation tools.

and yes, I know about the legalities in WiFi firmware but resolved now?

#agenticengineer #ai #llm #embedded

⚡ Create, edit and share videos at no cost in Google Vids

New AI capabilities are coming to Google Vids, powered by Lyria 3 and Veo 3.1, including high-quality video g…
#AI #MachineLearning #LLM
AI · blog.google/products-and-platforms/p...

The #AIBubble cannot burst soon enough.

Sure, #LLM systems will probably remain with us. But without the Bubble, the trillion dollar business case for "let's scrape every website a thousand times per month in the hopes of finding some new scrap of training data" should vanish. I hope...

Я не программист. Я два месяца учил нейросеть не подлизывать Нейросеть хвалила мой говнокод и помогала зако...

#BRO #Бро #промпт-инжиниринг #системный #промпт #LLM #нейросеть #Gemini #Unreal #Engine #галлюцинации

Origin | Interest | Match

The Donnie Darko "I made a friend" meme: - I built a tool - built or got a slop extruder to do it? - Slop…

The state of software "development" announcements in 2026 :NoAI:

#ai #llm