TrendPulse

One thing has felt odd to me while working with AI agents. We've gotten pretty good at giving them memory for text. They can search documentation, index repositories, retrieve past conversations, and even build long-term memory over time. Videos, though, are still treated as temporary input. The agent watches a recording, answers a few questions, and when the session ends, that understanding is usually gone. Next session, the same video gets processed all over again. That feels like an...

I have been asking Claude to give me feedback on dates I go out on with girls. Most recently I went on a really fun date that unfortunately ended with ghosting so I put as much detail as possible about the date into Claude Sonnet 5 and asked it for feedback. After running the date through Sonnet 5 several times it just kept repeating that it was a solid date and it did not have feedback to provide. I do not want validation, I want to improve to reach my goals. Has anyone found a workaround to...

ChatGPT is asking me to close ChatGPT... Which is honestly surprising to me, especially if it's just a way to reduce cost. Which I doubt will do anything If true..

https://orbis.aurochthryx.com

A clear trend is happening lately, a lot of users are moving away from heavily restricted models like chatgpt and claude toward uncensored or local models. Common reasons seem to be fewer refusals, better creative freedom, and privacy concerns. Has anyone else made the switch or considered it?

Seeing as there are about 1000 to choose from, I'm looking for one that just takes the prompt and runs with it. It doesn't have to be free, I'm going to be using this for business.

I was in the ai industry but realized my mind and ai could create something dangerous so now I’m hearing about all this is it worth it or what’s going on really in the industry

wanted to share a project i've been working on that explores using LLMs for social, in-feed entertainment. it's a text-based simulation game called Slop-Cops, built on reddit's new developer platform (devvit). users submit a website URL or describe a project idea (like an AI startup), and 5 distinct AI agent personalities (acting as a tribunal of "vibe cops") read the text, debate its quality, and rate it. players can then write a rebuttal to defend their project before the final ve...

Most of the focus is on AI being bad rather than how major companies are deploying AI. My concern isn't that AI is becoming more powerful. I mean, that is a concern, of course, but since most of the implications are speculative, you can't exactly take any stance or action on that problem other than countries coming together and setting rules and policies for how they distribute and use frontier models and capabilities, especially in warfare. My largest concern is what corporations and governme...

Anthropic has accused Alibaba of creating tens of thousands of fake Claude accounts to scrape Claude of its intellectual property via distillation attacks . Alibaba retaliates by telling their official (not contracted) employees to stop using Claude Code. I'm noticing from Reddit posts and comments that Claude has gotten much more wary of what it determines as strange prompting requests? There is an article indicating that Fable 5 has been "hardened" against distillation ...

When ChatGPT first came out, I assumed I'd mostly use it to answer random questions. That lasted about a week. Now the thing I use it for the most is taking messy thoughts and turning them into something I can actually work with. Whether it's rewriting an email, organizing notes, or helping me think through an idea, that's become the real value for me. Ironically, I use AI less for getting answers and more for helping me think more clearly. What about you? What's one use case you genui...

i run a few accounts at decent volume and the content production was tough what was helpfyl was connecting the claude mcp from a social media management tool. a lot of them are rolling this out right now you connect the mcp once, and after that you just chat with it. you build the content plan out in plain conversation, it generates the posts with consistent branding and a human sounding voice, and you approve everything from one place the stuff most people get wrong when they try to automate...

Devo dirlo, sono diventato matto...questa volta pensavo di mollare sul serio, ma poi stamattina il miracolo. (almeno per me). Ho fatto fine tuning e scritto tools per due modelli: un Qwen 3 da 1.5B e un altro da 4B quantizzati all'estremo. Il 4B è per smartphone con almeno 12Gb di RAM, mentre il piccolo è per la fascia media (occupa appena 2,5 GB di RAM)...sto creando un LoRa in bilingua per aiutare il piccolo a usare al meglio i tools, e lo sto distillando da un 32B, quindi meglio di uno fatto ...

It feels like most AI discussions revolve around AGI timelines or benchmark scores, but there are a lot of smaller capabilities improving rapidly (reasoning, memory, multimodal understanding, coding agents, robotics, speech, etc.) Which capability do you think is currently underrated and why? I'm more interested in practical applications over the next 3–5 years than distant predictions.

I ran the same market-entry brief through three different AI models. The result was revealing. I asked three models to independently create a client-ready market-entry brief for launching a privacy-first AI personal assistant for small businesses in the UK. The models were: Claude Fable 5 via Claude Subscription GPT-5.5 via ChatGPT/Codex qwen3.6:27b running locally via Ollama Each got the exact same task. They could use web research. They could not see each other’s answers. The...

Hey all ​I've been evaluating the "AI coworker" approach between the new Databricks Genie One interface and Anthropic's Claude Cowork desktop agent. While both claim to move past standard chat prompts into autonomous execution, they handle data from opposite ends of the stack. Claude Cowork operates right on your computer, meaning it's incredible for unblocking local workflow bottlenecks like organizing a messy folder of exports or synthesizing a bunch of text documents into a fresh ...

With AI advancing so quickly, I'm trying to understand how to prepare for the next 5–10 years. A few questions I'd love to hear your thoughts on: What skills should people focus on learning now to stay relevant alongside AI? Beyond prompting, what are the most valuable AI-related skills (automation, coding, workflows, AI agents, etc.)? Which jobs or industries are likely to benefit the most, and which are at the highest risk? Do you think AI will become expensive to use in the future? ...

I think the way most people use AI is stupid. They depend on the AI to do everything, even the thinking. They don’t even know the full capability of AI yet. They still think it’s talking to chat gpt on the app. I’ve had an ai agent installed via Hermes on my computer for a while now, and I’ll say this, ai only amplifies who you are, and unfortunately, most people are idiots or immoral or both. This is why there are “ai haters” what they’re really hating on is the shitty , lazily automated ai slo...

A month ago I posted the very rough beginnings of a paper. That rough version did not survive: it got pulled apart and rebuilt by the very process it describes, and what came out the other side is now a proper preprint with a DOI: https://doi.org/10.5281/zenodo.21139628 . Short version: the core claim held. The artifacts (specs, plans, executable graphs) and the verification gates wrapped around them have proven out on real work. Agents produce the work, the gates catch the defects, and a miles...

https://orbis.aurochthryx.com

As allegations of LLM use rock the literary and media worlds, linguists explain what really distinguishes human and machine writing, while novelists including Jennifer Egan and Jeanette Winterson reflect on the future of fiction in an age of ChatGPT

Saw a national survey from March on how people feel about AI in politics, and two numbers stuck with me. 63% said they'd be uncomfortable using an AI chatbot to help decide who to vote for, even though plenty of people are fine using chatbots to fact-check or follow issues. 80% said they're worried that AI bots, not real people, are answering the surveys that feed into policy and business decisions. It reads less like fear of the tech and more like people drawing a hard line at AI touch...

IM a nobody. So I worked my ass off and spend my last 800 $ (couldn’t pay my bills that month) but I created something that beated Microsoft and research teams on they benchmarks and open source it! It actually got a lot of stars on GitHub, some traction you know. Hoping to get some offer to get out of my box. But I was nervous, the was so good that I didn’t know what to do, when I saw the potential again, I made a mistake that cost me too much. I asked my claude (nerfed as a side character...

Hello everyone, There are many artificial intelligences on the market. There are the most well-known, but there are also others. I have an iPhone as well as an iMac, but today I am disappointed with ChatGPT and I am looking to replace ChatGPT with another artificial intelligence. I have no idea where to go. I do not know what would be, in your opinion, the best artificial intelligence if you know the prices, the advantages or other. I really need you.

There are a lot of web agencies doing email automation to land web design projects. They keep testing new email sequences every week, adding more follow ups, changing subject lines, and trying everything they can to increase their reply rate, but a lot of them still struggle. I was in the exact same position until I completely changed my strategy. The biggest change wasn't the sequence itself, it was the way I approached outreach. Instead of sending generic emails talking about my agency or as...

There's a handful of use cases that get pitched nonstop in demos and decks, and then completely fall apart the second you try running them for real. For me it's anything involving reading intent from ambiguous human input. When you give it a clear support ticket everything would be fine. But give it a message where the person's clearly annoyed but not saying why, and it either overreacts or misses it completely. What are those kinda use-cases for you? And it doesn't need to be some big drama...

Just signed up for Perch AI and I'm looking at Pro for $10/mo (says 50% off, "Early Access pricing"). Couldn't find anything on the pricing page or FAQ about how long this actually lasts. Has anyone been on Pro for a few months now? Did the price jump back to $20 at some point, or is it still $10 for you? Also curious if this is locked in per account once you subscribe, or if it can change mid subscription without warning. Trying to decide if it's worth committing to before finding...

Hi, I want to train a Transformer layer to perform specific tasks, but I’m not sure how to coordinate them or determine when to use one versus the other. Does anyone have experience with this? How have you handled it?