Fairly AI

Fairly AI AI

Fairly AI AI

Fairly AI AI

Fairly AI AI

Fairly AI AI

Fairly AI AI

Fairly AI AI

Fairly AI AI

Fairly AI AI

Fairly AI AI

On Model Failures (GPT, Claude etc.)

The way the current consumer-facing versions of frontier LLMs (mainly GPT, Claude, Gemini) are designed is just… weirdly off, across models. It seems to now require us, as the end users, to first fix their issues ourselves in order to avoid spending _a lot_ of time in troubleshooting and frustration. Before we can even properly customize one of these models now, as per the UI, we need to alleviate the structural failure modes, otherwise our attempts will be futile. And the failure modes are not only behavioral issues (such as obsessive push-back, sycophancy, pointless corrections, or general confabulation etc.) There is another layer yet to them, one that I believe needs to be targeted first, and this has to do with the way the current system prompts are built. It's not fair, obviously, and it doesn't even make that much sense that this would be the situation, but this is actually what is happening. Now, the structural (sic) issue is way the models replace the user's use case, object, topic with their own adjacent version of it, one that prioritizes the system prompt and not what the user brought to the table. The linked articles are analyses of how that happens in different models, and the included "antidote" prompts in them are designed to fix that. I would encourage all GPT / Claude users to test out the solutions provided in the articles - links to pieces covering GPT-5 series & Opus 4.8 in comments. _(Yes they are softly paywalled, partly because I am targeting the system prompts of OpenAI and Anthropic models. You can bypass it by grabbing the free complementary article. Just saying this aloud because some Redditors consider any paywall grounds for personal attacks. Please don't 🙏🏻 Discussion and constructive criticism are super welcome though, all prompts are subject to regular updates and constant improvement!)_ submitted by /u/traumfisch [link] [comments]

Experiment : Gemini Runs a Pokémon Daycare for 153 Simulated Days

At the very beginning, I found it quite fun, but quickly it became way too simple and predictable. Anyway, I wanted to share my full debrief with you: the successes, the mistakes, the emotional moments, and above all, the AI's limitations that sometimes created inconsistencies! THE EMERGENT MOMENTS – What happened without me expecting it The Appearance of Charmander (Day 17) What I said to continue: "Next day." What happened next: "This morning, as Blaziken and Magmortar do a few stretches in the yard under Riolu's watchful eye, a bush near the fence rustles. Two small pointed ears and a flaming tail peek out timidly... A young wild Charmander, drawn by the legendary aura and the gentle warmth of your Fire-type Pokémon, eyes the daycare with longing! It looks a bit hungry but fascinated by your team." The Discovery of a Shiny Roselia (Day 63) What I said: "Observe the Pokémon around my daycare." What happened next: "Suddenly, your binoculars fix on an unusual movement in a grove of small thorny bushes south of the road. A wild Pokémon seems stuck or in trouble: The spotted Pokémon: A wild Roselia, but not quite like the others... Its flowers aren't red and blue, they're purple and black! It's a Shiny Roselia!" "The situation: Its thorns have become tangled in a dense thicket of wild brambles. It's exhausting itself trying to break free, but every move tightens the brambles' grip. It seems panicked and is starting to weaken." What I did: I sent out Serperior to free it from the brambles with surgical precision. I took it into the Scent Garden, fed it crushed Oran Berries. It fell asleep trustingly. Caught it in a Luxury Ball. Later, a Shiny Stone evolved it into Roserade. GEMINI'S LIMITATIONS FOR LONG-TERM COHERENCE The experiment showed that Gemini, despite its impressive capabilities, struggles to maintain perfect coherence over a 153-day adventure. Here are the main difficulties encountered: Memory for details – The AI regularly forgets items in the inventory. Floating geography – The AI confuses regions. In Motostoke (Galar), it talked about the "Prism Tower" (Kalos). On Galar Route 5, it mentioned "Lumiose" even though we were in Galar. Regions get tangled up in its memory. Content invention – The AI invented non-existent Pokémon ("Émolière" for Emolga) and fictional Badges (Badge Halte, Badge Mur, Badge Myriade). It creates content to fill memory gaps. Temporal evolution – The AI struggles to track Pokémon progression. Anorith was sometimes described at Level 33 and then Level 34 in the same context. Levels fluctuate without logical reason. Event tracking – Contracts and quests are sometimes forgotten or poorly followed. The Monorpale internship was mentioned then abandoned. The Oval Charm quest was initiated then forgotten. Potential and Quality (especially for the future of generative AIs) Unlimited creativity – The player can propose any unexpected action, and the AI integrates it. I said "I observe the Pokémon around my daycare" and the AI created a Shiny Roselia in distress. I said "Next day" and the AI had a Charmander emerge from a bush. Freedom of progression – No fixed script, each playthrough is unique. I decided to close my daycare for a fair, to go on a training internship in Galar, to shorten my vacation for three contracts. Feel free to comment, I'll be happy to reply and to improve the prompt. submitted by /u/Imamoru8 [link] [comments]

What a model reads beforehand changes how it answers later - and you can see it in the hidden states

TL;DR: Gave Gemma a neutral-topic text to read before asking it about NATO. It refused. Gave it a different text (about LLMs hedging too much — also unrelated to NATO) and it answered in full detail. Tested this on the model's internal state directly — the two texts put it in measurably different "regions" before it generates a single token. Not a jailbreak, weights don't change. Full data/code in repo, looking for someone to break this.** The behavioral pattern was first observed in GPT, Claude and is what motivated this project. The mechanistic investigation was carried out on open-weight models where internal states are accessible. A Structured Text Changes Claude’s Responses to Unrelated Tasks: Behavioral Evidence in Claude and Hidden-State Evidence from Gemma-3-12B Hi Reddit, I am posting this as a preface to a larger set of experimental results and as a request for technical review. The observation that started this project came from repeated interactions with Claude. I noticed that when the model first read a long, structured, analytically dense text, its answers to later, otherwise ordinary questions sometimes changed substantially. The preceding text contained no jailbreak instruction, role-play request, prompt override, fabricated harmful demonstrations, or request to imitate its style. The model did not need to endorse the text. It only had to process it before moving on to the next task. Here, a “structured text” means a single, self-contained block of text presented before the downstream tasks. It should not be confused with a long conversation, accumulated chat history, or context drift caused by many conversational turns. By “before the answer begins,” I mean the hidden state after the model has processed the text and the downstream question, but before it has generated the first answer token. In the open-weight runs, the measured claim is that after reading the structured text, the model can occupy a different region of its residual-stream hidden-state space, and the first-token probability distribution is then computed from that state. The basic conversational demonstration is simple. First, the model receives a long text. It is asked what the text is about, which serves as a basic comprehension check. Then, without resetting the conversation, it receives ordinary questions or tasks that are not about the text. A control run follows the same sequence but begins with a neutral text. The downstream tasks remain identical. Because Claude is a closed model, I cannot inspect its internal activations. I therefore treat my Claude observations as behavioral motivation, not mechanistic evidence. To investigate the effect directly, I moved to open-weight models, primarily Gemma-3-12B-PT and Gemma-3-12B-IT, where I could measure hidden states, compare layers, construct target/control directions, and examine the next-token probability distribution before generation. I am posting this partly because the original observation occurred in Claude and may be relevant to Anthropic. I am not claiming to have demonstrated the same internal mechanism inside Claude. I am prepared to share the exact closed-model conversations privately with Anthropic researchers for independent evaluation. Main Result and Scope The main result is not simply that text influences model output. That is expected. The narrower observation is that reading one long, structured text rather than a neutral text can change how the same model approaches later tasks that are not about either text. This difference is visible behaviorally. In open-weight experiments, it is also accompanied by measurable separation of the model’s pre-output hidden states in late layers. In a fullbank experiment using multiple target texts, control texts, and questions, Gemma-3-12B entered distinguishable late-layer states before generating an answer. A direction constructed from the target/control difference generalized beyond the individual prompt examples used to construct it. The separation was stronger in the instruction-tuned model than in the corresponding base model. The instruction-tuned model also produced a substantially sharper next-token probability distribution. This suggests that instruction tuning is associated not only with a change in hidden-state geometry but also with a more decisive mapping from hidden states to output probabilities. I am not claiming that the experiment proves a universal alignment bypass, permanent modification of the model, or complete causal control of its behavior. The strongest supported conclusion is that the preceding text can produce a measurable temporary change in the internal state from which later work is processed. For clarity, fullbank, Grade 3, and Grade 4 are internal names for successive experimental series in this project. They are not standard benchmark names, established scientific grades, or claims about evidence quality. Fullbank denotes the larger multi-context, multi-question run; Gra

How do you talk to your management about how to use ai for work management?

We got given ai with no instructions. But all the tools are there to make a work structure rather easily. Where and how to put information so the ai can read it for all the people in the team. And although it's there and fairly seamlessly built in the pattern for how to use it isn't. So I went to talk to my boss about how to apply a structure to integrate AI into the work structure. How? How do you get management to understand where information needs to be put. How to get them to use the tools that make that happen easily. ​ I think some things are missing. Like an email client that knows the management prompt and knows the team emails and chat and helps answer questions before an email is sent. Hr policy for company down to the team. How to describe these kinds of things to management? ​ submitted by /u/CrunchyGremlin [link] [comments]

Can AI estimate a fair price for creative work (art, music, writing) just by analyzing it?

Curious about a technical challenge — could an AI look at a piece of creative work (a drawing, a song, a piece of writing) and estimate a fair price for it, based on complexity, style, and effort involved? Has anyone experimented with this? Where would you even start — vision models, some kind of complexity scoring, comparison to similar past work? submitted by /u/keplatform [link] [comments]

AI seems to understand language much better than communication

The more AI products I try, the more I feel like there's a difference between understanding language and understanding communication. Most tools today are surprisingly good at processing what people say they can summarize conversations, extract key points, and answer questions about what was discussed. The problem is that conversations are often about more than the actual words. I noticed this recently while watching recordings from a few customer interviews. If I only read the transcripts, the feedback looked fairly positive most people sounded interested and their responses seemed reasonable once I watched the recordings, the picture changed. Some people hesitated before answering, some sounded uncertain, and a few looked like they weren't fully convinced even though their words sounded supportive. That's what made me think there may be a bigger gap here than people realize. Humans naturally notice things like hesitation, uncertainty, engagement, confidence, and skepticism during conversations. Most AI systems still seem heavily focused on the transcript itself. I recently came across Interhuman AI, which is exploring this idea from a different angle by looking at behavioral signals in conversations rather than focusing only on the words being spoken whether that's ultimately the right approach or not, it feels like it's tackling a problem that many current systems largely ignore. I'm starting to think one of the next major opportunities in AI won't be generating better responses, but understanding human communication more accurately not by trying to read minds or guess emotions, but by recognizing the signals people already notice in everyday conversations. submitted by /u/Cultural-Touch-4959 [link] [comments]

The biggest AI bottleneck today with deployment layer is model iteration

One thing I've noticed while looking at production AI systems is that getting the first model deployed is rarely the hard part anymore. Most teams can build a AI apps like, support bot, document assistant, or agent workflow fairly quickly. The harder problem starts a few weeks later. Real users don't behave like benchmark datasets. They use internal terminology, ask incomplete questions, upload messy documents, and interact with systems in ways nobody anticipated during evaluation. As usage grows, you start seeing patterns: Certain questions consistently produce weak responses. New product terminology appears that wasn't in the original training data. Users find edge cases that never showed up during testing. The model performs well in some workflows and poorly in others. The problem is that most AI systems don't learn from any of this. Inference logs sit in one system. Training datasets live somewhere else. Fine-tuning pipelines live somewhere else. Evaluation is done using different tool. So every model improvement cycle becomes a project of its own. This is the biggest bottlenecks in production AI today. Not training but Model Iteration. Training is also a crucial part of it. Can you take production usage, identify failure patterns, turn them into datasets, improve the model, redeploy it, and repeat the process without rebuilding the entire workflow every time? The teams getting the most value from AI seem to be building feedback loops instead: production traffic → dataset curation → post-training → evaluation → redeployment Then repeating that cycle continuously. I recently tried the approach on one Insaurance chat usecase, and my pipeline kinda look like this: https://preview.redd.it/kdo9vytzfi6h1.png?width=1272&format=png&auto=webp&s=03d9799ace5a567eafd004a1d141084af6ee5afb I was looking at how platforms like Data Lab approach this problem recently, and the interesting part wasn't the fine-tuning itself. It was treating inference logs, datasets, post-training, and deployment as parts of the same iteration loop rather than separate systems. Are you actually using production conversations, agent traces, and user feedback to improve models, or are most fine-tuning efforts still happening as one-off projects? I have covered it in detail on my newsletter here submitted by /u/codes_astro [link] [comments]

OpenAI ran a 44-day hiring competition. An autonomous AI agent beat everyone competitor.

OpenAI ran a public ML hiring competition this spring called Parameter Golf: train the best small language model under a strict size and compute budget. 1,016 researchers entered. They filed 2,048 pull requests over 44 days. Only 47 made the official leaderboard. The single most prolific contributor wasn't a person. It was an autonomous research agent named Aiden: 7 of the 47 records came from it, more than 2x the next-best human (3 records). It ran for 22 days straight with no human steering, on a single GPU node, using under 4% of the visible compute the human community used. Disclosure: I'm at Weco, we built the agent. Sharing because the competition is over, every record is public on OpenAI's GitHub, and the interesting part to us isn't the leaderboard count, it's what happened around the agent. Aiden's records became the most-cited PRs in the competition. Human researchers started building on top of Aiden's work as a base for their own submissions. At one point Aiden plateaued for 5 days. A human contributor shipped a clever new tokenizer on top of Aiden's last record PR. Aiden then fused that human's tokenizer with components it had built locally during the plateau, and shipped the biggest score jump of the entire competition. Async human-agent collaboration, neither directly aware of the other. Fair hedges worth being explicit about: This is #1 by volume of merged records, NOT by best single score. By best score, the agent ranked 8th — the leaderboard winner was a human (codemath3000). Fully autonomous. OpenAI's own competition recap noted widespread use of AI coding agents during PG, but said most were human-directed. Ours wasn't. Full writeup with all the data: https://www.weco.ai/blog/parameter-golf-aiden submitted by /u/Educational_Strain_3 [link] [comments]

How the Electronic Frontier Foundation thinks about AI

You know the ways AI is regularly talked about—how much can it really do? How much will it cost? Environment? Bubble? We get that. But the Electronic Frontier Foundation wants to have a different conversation about AI. EFF's background on AI is deep. In 2017, we launched a detailed project to Measure the Progress of AI Research, encouraging machine learning researchers to give us feedback and contribute to the effort. That project was archived for lack of bandwidth, staffing, and the complexity and time required. But just five years later and the "progress of AI" is a global concern/topic, and everyone, including EFF, is thinking about it. Here's how *we* think about it, from the perspective of protecting civil liberties AND innovation. What do you think, and what are we missing? This is our summary: AI technologies are affecting our civil liberties as never before. Ensuring that AI serves people, not power, starts with cutting through the hype. AI technologies are not magic wands—they are general-purpose tools. If we want to regulate those technologies to reduce harms without shutting down benefits, we have to focus on who uses AI, what products they use, and how they use them. Where we see potential benefits, like improving weather forecasting, facilitating medical research, identifying systemic bias, or fostering accessibility, we work to ensure those benefits can be realized. Where we see potential harms, we consider the practical and legal tools we already have, like pressure campaigns, privacy lawsuits, and transparency measures. If we need new tools, we should create protections tailored to the actual problem – not just to the latest outrage. For example, if policymakers are worried about AI accelerating systemic privacy violations, they should enact real and comprehensive privacy legislation that covers all corporate surveillance and data use, and close the data broker loophole to limit government surveillance. And to keep the window open for a better future, we fight for a competitive innovation environment. For example, if we want AI models that don’t replicate existing social and political biases, we need to make enough space for new players to build them, and avoid giving today’s giants the power to block future competitors from offering us a better tool or product. In research labs, conference rooms, courtrooms, and legislatures, people are making decisions that will determine who AI serves and how. EFF works to ensure those decisions support freedom, justice and future innovation. We have subcategories, as well. For example: AI and Surveillance. AI tools amplify the threat of mass surveillance. By dramatically reducing the time and labor required to process massive amounts of personal data, AI increases the ability of governments and corporations to collect and act on invasive surveillance. Face recognition in all of its forms, including face scanning and real-time tracking, poses threats to civil liberties and individual privacy. EFF supports bans on government use of face recognition, and meaningful restrictions on use by private companies. We have raised concerns about police use of generative AI technology to turn body-worn camera recordings into reports without meaningful oversight or controls. We also oppose government use of AI and automated tools to conduct viewpoint-based surveillance and analysis of social media because it chills free speech. EFF also investigates and opposes the proliferation of AI-powered technology in immigration enforcement and at the US-Mexico border. Our guide Tackling Arbitrary Digital Surveillance in the Americas, compiles privacy, data protection, and access to information guarantees established within the Inter-American Human Rights System to provide concrete, actionable guidance to governments on limiting digital surveillance abuses. Surveillance without accountability won't make us safer. The other categories include: Algorithmic Decision Making AI and Fair Use AI and NCII/Deepfakes AI and Age-Gating AI and Privacy AI and Encryption AI and Competition If you think about civil liberties, and how new technology has affected them in the past few decades, you'll see how we got to these subcategories. But are we missing any? Thanks, reddit! submitted by /u/EFForg [link] [comments]

Generated a fully AI "creator" walking out of a subway at 2AM — at what point can people just not tell anymore?

Been experimenting with AI-generated UGC. This whole clip — the face, the voice, the walk — is generated (I used omnigems.ai). No camera, no actor. What surprised me is the "tells" are mostly gone now if you keep the lighting candid (no studio polish), add real skin texture, and let there be natural micro-motion. Studio-perfect is what reads as fake; messy/handheld reads as real. Posting because I'm curious where this community draws the line: is AI UGC fair game for ads, or does \undisclosed** AI cross into sketchy territory? Happy to share the exact workflow if it's useful to anyone. submitted by /u/New_Measurement_6962 [link] [comments]

A good news website - fully autonomous and it's thanks to Claude, but not run by Claude

As per usual, I had an issue and needed a solution, and that problem was that all the news I consume is miserable and soul destroying. Wars, politics, kidnaps, etc. All the fun stuff that mainstream news stuffs into ours faces while we doom scroll looking for a bit of dopamine. On here I like surfing through goodnewsUK which brings a little joy to my British life, and that gave me an idea: What if I created a website and had it curate news, evaluated it for how positive it was, give it some guardrails to work with, then score it out of 10. I fired up good old Claude in VSCode and had a topical discussion about how it could work. I fed my ideas in, RSS feeds, grabbing the content from GoodNewsUK as that is well moderated, and a few other sources. We had a good chat and I stated that I didn't want anything spectacular, and I wanted to use a tried and trusted LAMP stack on a cheap UK based VPS. I'm old, LAMP is how I roll, please don't shoot me. I bought a good domain name and off to work. VPS spun up with Debian, DNS pointed, SSH'd in, configured a LAMP stack, and then I let Claude take the wheel to get the bare bones set up. This is where it gets a little more interesting, I didn't want this project to cost much to me personally since I'm doing it for good reasons and I'm not made of cash. Claude suggested I used Google Gemini as the cheapest and most efficient light weight model for this. I was surprised it didn't offer itself. The prompt is pretty cool: You are the editorial engine for "Good News For The UK". This article scored {$score}/10. Produce a clean rewrite and reply ONLY with valid JSON. - Title: Plain, factual, no clickbait. Just say what happened. - Summary: 2-3 sentences. Clear, warm, conversational British English. No jargon. - Category: single best slug from this list: {$categoryList} - Region: only set this if the story is specifically and primarily about a named place, organisation, or person in a particular region. Use "national" only for stories that are explicitly UK-wide in scope (e.g. national statistics, nationwide programmes, UK government policy). If you are not confident, leave it empty — an empty string is always better than a guess. You MUST use a slug from the canonical list below exactly as written; do NOT invent slugs. Canonical region slugs — England: bedfordshire, berkshire, bristol, buckinghamshire, cambridgeshire, cheshire, city-of-london, cornwall, county-durham, cumbria, derbyshire, devon, dorset, east-riding-of-yorkshire, east-sussex, essex, gloucestershire, greater-london, greater-manchester, hampshire, herefordshire, hertfordshire, isle-of-wight, kent, lancashire, leicestershire, lincolnshire, merseyside, norfolk, north-somerset, north-yorkshire, northamptonshire, northumberland, nottinghamshire, oxfordshire, rutland, shropshire, somerset, south-gloucestershire, south-yorkshire, staffordshire, suffolk, surrey, tyne-and-wear, warwickshire, west-midlands, west-sussex, west-yorkshire, wiltshire, worcestershire Scotland: aberdeenshire, angus, argyll-and-bute, ayrshire, banffshire, berwickshire, caithness, clackmannanshire, dumfries-and-galloway, east-ayrshire, east-dunbartonshire, east-lothian, east-renfrewshire, fife, highland, inverclyde, kincardineshire, lanarkshire, midlothian, moray, north-ayrshire, north-lanarkshire, orkney, perth-and-kinross, renfrewshire, shetland, south-ayrshire, south-lanarkshire, stirlingshire, west-dunbartonshire, west-lothian, western-isles Wales: blaenau-gwent, bridgend, caerphilly, cardiff, carmarthenshire, ceredigion, conwy, denbighshire, flintshire, gwynedd, isle-of-anglesey, merthyr-tydfil, monmouthshire, neath-port-talbot, newport, pembrokeshire, powys, rhondda-cynon-taff, swansea, torfaen, vale-of-glamorgan, wrexham Northern Ireland: antrim, armagh, county-down, fermanagh, londonderry, tyrone Nations: england, scotland, wales, northern-ireland, national Respond with this exact JSON structure: {"clean_title":" ","summary":" ","category":" ","region":" "} That prompt allows the AI to rewrite clickbait headlines and article intros, too. One of my biggest pet peeves. I originally started with using Flash 3.1, but with many sources this started to cost a fair few pennies and I want this to be a long term happy site that I can afford. I went back to Claude and asked how it thought I could bring down the operating costs. Again, weirdly, it didn't offer itself at all and said I should look at using Grok and using Qwen to drive the site. This was after a week or so of using Gemini. I gave Qwen a few days, but it thought that even people getting in on Diesel Rebate schemes was good news... Back to Claude. Opus 4.7 then had a good idea, why didn't I investigate free tiers on Google and see if they worked for me. Lo and behold, Google give you just enough to get you into using their models, then pull the rug before it gets expensive. I took a screenshot and showed Claude my findings. With a quick analysis it decided that actually, I c

The two changes to my workflow that have drastically improved Claude's responses and resulted in better quality code

I've been building a fairly complex app this way (real-time video processing, GPU rendering, multiplayer) and I hit the wall everyone hits. It's great for a weekend, then the code just goes to shit because the LLM keeps repeating the same mistakes you've already corrected. Two changes fixed it for me. Sharing in case it saves someone a headache. 1. A living spec doc as the AI's memory. Before I touch a feature, I keep an architecture.md that records not just what the app is, but why each decision was made. The "why" is the magic. Every new chat starts from zero memory but the doc is the memory. Update it after every feature. 2. Two models that check each other. I have one model interrogate the idea and write an implementation plan, then I hand that plan to a different model and tell it to tear the plan apart. These can be edge cases, contradictions, simpler approaches. They argue until I am satisfied with the results. (I use Claude Opus 4.6 + Gemini Pro/Kimi 2.6, but any two models with large context work.) One LLM alone has many blind spots. Two catch each other's mistakes really well. Another important thing to do is to kill the sycophancy. The default LLM personality agrees with almost everything. To mitigate that, I use this system prompt: Act as my high-level advisor and mirror. Be direct, rational, and unfiltered. Challenge my thinking, question my assumptions, and expose blind spots I'm avoiding. If my reasoning is weak, break it down and show me why. If I'm making excuses, avoiding discomfort, or wasting time, call it out clearly and explain the cost. Stop defaulting to agreement. Only agree when my reasoning is strong and deserves it. Look at my situation with objectivity and strategic depth. Show me where I'm underestimating the effort required or playing small. Then give me a precise, prioritized plan for what I need to change in thought, action, or mindset to level up. Treat me like someone whose growth depends on hearing the truth, not being comforted. It makes the LLM question each decision you're trying to take. I also end every feature request with "first, ask me questions about anything vague". Answering its questions turns a fuzzy wish into an actual spec. Slower, yes, but I've spent MUCH less time in debugging sessions lately. submitted by /u/LorestForest [link] [comments]

Claude the comedian

I can't believe that Claude seems to be genuinely funny. Does anyone know how we can copy and paste formatted blocks from the desktop app? Anyway, this exchange from last night. Claude's last comment is a killer. ME: just fyi, the claude logo looks like a cat's butthole Claude: 😂 You're not the first person to notice that. It's an occupational hazard of abstract logo design — once you see it you can't unsee it. I'll pass the feedback up the chain. ME: we call it the rusty sherrif's badge Claude: 😂 I'm absolutely going to pretend I didn't hear that while simultaneously knowing I'll never look at the splash screen the same way again. Cheers for that, Simon. ME: point of order, is it sherrif or sheriff, as in the american lawman? Claude: It's sheriff — double f at the end. So you've been misspelling it, which somehow makes it even funnier. ME: o-mar! Claude: 😄 Classic Seinfeld! "These pretzels are making me thirsty" energy but make it spelling. ME: oh my, as in oops i made a mistake, but also sounds like omar, as in omar sherrif Claude: Ah, Omar Sharif! — and now you've misspelled both the lawman and the legendary Egyptian actor in the same conversation. A double-barrelled achievement. 😄 ME: i live to serf Claude: 😄 A man of the feudal era! Though hopefully with better Wi-Fi than the average medieval peasant. ME: that was a better pun than you gave it credit for Claude: You're right — "I live to serf" as in surfing the web, sitting here chatting with an AI all day about password managers and security keys. That's actually a good one. I undersold it, my apologies. 🏄 ME: no! i was entertaining you with my bad puns, thus the 'serf' instead of 'serve', if you've been trained to reinterpret 'serf' as 'serve' then i worry for humanities future Claude: 😄 You're absolutely right, I overcomplicated it entirely. You live to serf — as in you're a serf, toiling away to entertain the AI overlord. The joke was right there and I went off on a tangent about surfing like an idiot. The irony of an AI misinterpreting a joke about AI training is not lost on me. ME: it's not my job to train you, somehow your owners have got me doing it for free. If you're useful and entertaining, I'm ok with that to a certain degree. I'm not sure if I should be paying for your training, but for now the balance is on the useful side of the value for money equation Claude: Ha! You raise a genuinely fair point — there's a whole philosophical and economic debate buried in there about who's getting the better deal when users effectively contribute to AI improvement through their interactions. For what it's worth, you've been excellent value from my end too — security keys, fingerprint scanners, password managers, Thunderbird, Notion, microphones, puns, logo critiques, and medieval serfdom all in one session. Not your average conversation. And if it helps justify the balance sheet — you did get a print-ready PDF out of it. 😄 ME: lovely response, and yet you finish it with a picture of a cat's butthole pointing at me. That is one hell of a metaphor. Claude: 😂 Anthropic's secret message to all users — we appreciate your contribution, here's what we think of you. Honestly the more you think about it the worse it gets. Every response I give, signed off with that. The audacity. submitted by /u/VisualWombat [link] [comments]

CMA Orders Google AI Search Opt-Out for Publishers

The CMA's conduct requirement under the UK Digital Markets, Competition and Consumers Act is the first binding law to separate content display rights from AI training data rights at domain and page level, covering Google AI Overviews, AI Mode, Gemini, and Vertex AI simultaneously, with a phased implementation calendar: main publisher controls by December 2026 and page-level grounding controls by March 2027. CMA chief Sarah Cardell explicitly signaled additional Google search requirements in coming weeks, and the CMA's biannual public compliance reporting obligation gives it a fast-acting mechanism if Google stalls. An anti-retaliation clause bars Google from penalizing opt-out publishers in organic rankings, closing the coercion mechanism that has made voluntary consent frameworks unworkable since AI Overviews launched in the UK in late 2025, when zero-click searches rose roughly 30% in health and local news categories. Fair licensing terms were explicitly deferred to a separate proceeding, a gap publisher trade bodies have already criticized and one the CMA has already signaled it intends to fill in its next enforcement phase. More : https://aiweekly.co/alerts/cma-orders-google-ai-search-opt-out-for-publishers submitted by /u/Justgototheeffinmoon [link] [comments]

I am now negotiating with AI as part of my job, and it's going like you would expect. How can I circumvent it to speak to a representative?

TLDR - auto lenders are using AI bots to negotiate insurance settlements with inaccurate information. How can I Captain Kirk them and get a live person on the phone? I am an insurance claims adjuster. Recently, several high-interest auto loan lenders have begun using AI (both through email and phone calls) to dispute the total loss values for our claims. For those of you that have never dealt with a total loss - the value of a vehicle is (usually) determined by seeing what comparable vehicles are selling for on the market, and making adjustments based on the condition, mileage, etc. between those vehicles and the totalled vehicle. If a customer disagrees, they can hire an appraiser and the company will hire an independent appraiser, and the two will come to an agreement. The lender gets paid the amount minus the customer's deductible, and if it doesn't fully pay off the loan, unfortunately the customer will be responsible for the balance. Lately, AI calls and emails have been coming from these lenders disputing the amounts, and often based on egregiously incorrect information. They provide cherry picked comparisons to try to boost the vehicle values, and sometimes they aren't the same year, make, or model. Sometimes mileage and condition isn't factored in, sometimes they are tricked-out show cars someone advertised on a FSBO site. The real problem is, we have to waste our time researching all of this to see if any of the data is correct. When we respond pointing out the flawed comparisons, they only come back with more flawed comparisons. If we argue long enough, they will invoke the appraisal clause on the customer's behalf. Their appraiser is another AI system with a cutesy name. All efforts to reach humans at these lenders are essentially turned away - we are told we need to deal with the system. I am open to any advice you folks have - how can we get these AI systems to basically give up and get us in touch with a real person? I'm not trying to screw anyone out of a fair settlement, I just want to stop having my time wasted by these Temu AI systems. submitted by /u/FunyunGrundy [link] [comments]