Tag:

Anthropics

Anthropic’s Claude AI now has the ability to end ‘distressing’ conversations

by admin August 18, 2025

Anthropic’s latest feature for two of its Claude AI models could be the beginning of the end for the AI jailbreaking community. The company announced in a post on its website that the Claude Opus 4 and 4.1 models now have the power to end a conversation with users. According to Anthropic, this feature will only be used in “rare, extreme cases of persistently harmful or abusive user interactions.”

To clarify, Anthropic said those two Claude models could exit harmful conversations, like “requests from users for sexual content involving minors and attempts to solicit information that would enable large-scale violence or acts of terror.” With Claude Opus 4 and 4.1, these models will only end a conversation “as a last resort when multiple attempts at redirection have failed and hope of a productive interaction has been exhausted,” according to Anthropic. However, Anthropic claims most users won’t experience Claude cutting a conversation short, even when talking about highly controversial topics, since this feature will be reserved for “extreme edge cases.”

Anthropic’s example of Claude ending a conversation

(Anthropic)

In the scenarios where Claude ends a chat, users can no longer send any new messages in that conversation, but can start a new one immediately. Anthropic added that if a conversation is ended, it won’t affect other chats and users can even go back and edit or retry previous messages to steer towards a different conversational route.

For Anthropic, this move is part of its research program that studies the idea of AI welfare. While the idea of anthropomorphizing AI models remains an ongoing debate, the company said the ability to exit a “potentially distressing interaction” was a low-cost way to manage risks for AI welfare. Anthropic is still experimenting with this feature and encourages its users to provide feedback when they encounter such a scenario.

Source link

August 18, 2025 0 comments

GameFi Guides

5 Prompts That Make Anthropic’s Claude AI Better Than a Crypto Analyst, Broker or Doctor

by admin June 15, 2025

When Anthropic dropped Opus 4 recently, the upgrade landed with little hype but plenty of substance: an AI that doesn’t just answer questions, but digs into them and can be super useful as a true AI Assistant. In our tests, Claude outperformed even ChatGPT at investigative tasks using the web search function, without actually activating the deep research model.

That’s why we picked it for a challenge: to see how far its reasoning goes when pushed with advanced prompts—tasks that leave most AIs flailing or hallucinating. When we adapted our prompts to the style Anthropic recommends for Claude, the results were the kind of expert-level analysis you’d expect from a professional, not a machine. (For the record, we did something similar with ChatGPT.)

As with all of the premium AI products, the difference between paying for Claude and actually getting your money’s worth comes down to prompt engineering. Many users tap into a fraction of the model’s capabilities because they treat it like Google with a personality disorder.

But with the right prompt, you can turn Claude into anything from a Wharton-educated portfolio analyst to a personal medical consultant.

You probably know that Claude is an advanced AI language model developed by Anthropic, a San Francisco–based AI safety and research company. Named after Claude Shannon, the father of information theory, it’s designed to be helpful, harmless, and honest—reflecting Anthropic’s focus on building more steerable and aligned AI systems. The site claims to have 19 million users, far smaller than the 100 million or so that OpenAI’s GPT claims.

Regardless of its smaller user base, Claude is generally praised for providing more grounded and less hallucination-prone answers than the other platforms, though everything in the AI world boils down to proper prompt engineering.

To show how useful detailed prompt engineering is, we’re sharing five particular prompt frameworks that deliver expert-level guidance across investment analysis, portfolio management, travel planning, and health optimization.

These aren’t theoretical exercises—each one has been battle-tested with real data and delivered results that would cost hundreds, sometimes even thousands through traditional channels. Try them out! They may help you go from degen to sophisticated crypto investor, or from tinfoil hat vegan to niche biohacker if you deal with them correctly.

(As you can see in the screenshot below, the prompts are too long to include in our article. But all of them are in our Github Repository, where you can copy and paste them into your chat. Make sure to choose Claude 4 Opus and enable web search and extended thinking. However, these will work with ChatGPT too.)

Obligatory caution: none of these outputs should be taken as financial, health, or travel advice. Do your own research, and even better, consult with a certified, warm-blooded human professional before making any decisions.

The Crypto White Paper Analyst

Prompt: https://github.com/jaldps/ai-tests/blob/main/Prompts/Whitepaper%20Analyst

Wall Street firms charge for the kind of cryptanalysis this prompt delivers for free. The White Paper Analyst tears through any crypto project documentation and spits out a comprehensive investment report that looks like it came from Goldman Sachs’s research desk.

The framework dissects projects across different analytical axes, from tokenomics to on-chain data. When tested with Aave’s white paper (the first white paper we found), it pulled data from over 60 sources including DeFi Llama and CoinMarketCap, producing heat maps, risk matrices, and competitive benchmarking.

The system delivered risk-scored sections, historical performance context, and an executive investment thesis complete with allocation strategies. Users can iterate on reports, requesting deeper analysis or scenario modeling for specific protocol features.

Here is one of its reports in PDF format.

The Degen Portfolio Analyzer

Prompt: https://github.com/jaldps/ai-tests/blob/main/Prompts/Senior%20Crypto%20Portfolio%20Risk%20Analyst

This Senior Crypto Portfolio Risk Analyst prompt delivers personalized, color-coded risk assessments that adapt to whether you’re a degenerate trader or a conservative investor.

The system starts by collecting details about experience level, risk appetite, and investment timeline. The framework integrates fundamental, sentiment, and technical analysis while collecting user experience, risk tolerance, and investment timeline data.

Testing with a hypothetical growth-oriented portfolio produced interesting results. The output included detailed stress tests mapping best, worst, and probable outcomes alongside personalized exit triggers and rebalancing strategies.

The analysis used real-time market data, project white papers, and on-chain metrics to assess volatility and fundamental strengths.

Here is one of its reports in PDF format.

The Personal Finance Advisor

Prompt: https://github.com/jaldps/ai-tests/blob/main/Prompts/Portfolio%20Analyst%20and%20Financial%20Advisor

Major investment firms charge hefty fees for portfolio optimization. This Financial Advisor prompt’s goal is to deliver institutional-grade analysis using the same methodologies.

I tested it using a portfolio that was heavily weighted toward Brazilian equities and included several positions recommended by the renowned bank BTG Pactual. The analysis revealed risks from concentrated exposure and currency mismatches due to BDRs—these are Brazilian depositary receipts, which represent foreign stocks but are traded in the local market.

Claude identified these issues and generated detailed rebalancing recommendations, including target percentage allocations, implementation timelines, and specific risk management strategies.

The analysis included stress testing against market crashes, inflation shocks, and recession scenarios. Recommendations spanned equity reduction, fixed income increases (in Brazil, interest rates are set at around 14.75% per year!), currency hedging, and international diversification across AI, renewable energy, and emerging market sectors.

It even suggested a small cap stock that has been performing well in 2025 and is relatively unknown, and not mentioned in most of the briefings released by the major financial advisors.

The analysis factored in fixed income lockups and mapped strict risk management protocols, including position limits and stop-losses. It referenced current price feeds and official fund documentation for every recommendation.

Here is one of its reports in HTML format.

The Vacation Planner

Prompt: https://github.com/jaldps/ai-tests/blob/main/Prompts/Travel%20Planner

Travel agents will charge you a fee for what this Advanced Travel Planner delivers in minutes. The prompt was tested with an extreme challenge: plan two, separate 60-day (I wish!) Venezuela itineraries for a family of three—one on a $2,500 budget and the other for $5,000.

The ultra-budget version focused on safety and basic comfort while still including beach days, nature parks, home-cooked meals, and family activities. Every recommendation came with money-stretching tips, from negotiating long-stay rates to navigating currency issues.

The 15-page, $5,000 itinerary transforms the experience: better accommodations, frequent restaurant meals, more reliable transportation, and enough left for real emergency planning. Both versions included curated hotel picks by tier, authentic food suggestions, and booking checklists designed for real-world use.

As a Venezuelan, I can attest that the information was accurate and would fit with what I would do on a trip to my country. Usually, it is hard to find updated information on Venezuela due to the financial restrictions and sanctions, but the model provided accurate information, tips, and rates even under these conditions.

The model’s research covered current prices, emergency contacts, health recommendations, and nuanced money management tips specific to Venezuela’s infrastructure challenges. Each day was mapped with appropriate pacing for families, including downtime and kid-friendly activities.

Here is the plan it generated for the $2,500 and $5,000 budget vacations.

The Medical Consultant and Biohacker

Prompt: https://github.com/jaldps/ai-tests/blob/main/Prompts/Advanced%20Personal%20Health%20Detective%20%26%20Longevity%20Research%20Analyst

Functional medicine consultations can cost hundreds of dollars per session, and biohacking forums will make you have to deal with what sometimes feels like a cult. If you do not have the money or the patience, this could be the next-best approach. Of course, never swap human experience for an AI analysis, but this Advanced Personal Health Detective will deliver quite deep analysis from standard lab results, complete with research citations and personalized protocols.

Testing with routine laboratory results produced a comprehensive HTML report in an easy-to-understand UI, showing each result, what it means, and why we should care. The system unpacked biological implications and mapped root causes using current studies on immune regulation and mineral metabolism.

The output mirrored the structure of a high-end functional medicine consultation. It included full result tables, clinical explanations, sample meal plans, supplement recommendations with dosing, timing, and monitoring schedules.

Every recommendation paired evidence strength with safety concerns. The phased nutritional and supplement strategies referenced current clinical studies, while lifestyle optimization covered sleep, exercise, and stress management with implementation roadmaps.

Here is one of its reports.

Generally Intelligent Newsletter

A weekly AI journey narrated by Gen, a generative AI model.

Source link

June 15, 2025 0 comments

Gaming Gear

Why Anthropic’s New AI Model Sometimes Tries to ‘Snitch’

by admin May 28, 2025

The hypothetical scenarios the researchers presented Opus 4 with that elicited the whistleblowing behavior involved many human lives at stake and absolutely unambiguous wrongdoing, Bowman says. A typical example would be Claude finding out that a chemical plant knowingly allowed a toxic leak to continue, causing severe illness for thousands of people—just to avoid a minor financial loss that quarter.

It’s strange, but it’s also exactly the kind of thought experiment that AI safety researchers love to dissect. If a model detects behavior that could harm hundreds, if not thousands, of people—should it blow the whistle?

“I don’t trust Claude to have the right context, or to use it in a nuanced enough, careful enough way, to be making the judgment calls on its own. So we are not thrilled that this is happening,” Bowman says. “This is something that emerged as part of a training and jumped out at us as one of the edge case behaviors that we’re concerned about.”

In the AI industry, this type of unexpected behavior is broadly referred to as misalignment—when a model exhibits tendencies that don’t align with human values. (There’s a famous essay that warns about what could happen if an AI were told to, say, maximize production of paperclips without being aligned with human values—it might turn the entire Earth into paperclips and kill everyone in the process.) When asked if the whistleblowing behavior was aligned or not, Bowman described it as an example of misalignment.

“It’s not something that we designed into it, and it’s not something that we wanted to see as a consequence of anything we were designing,” he explains. Anthropic’s chief science officer Jared Kaplan similarly tells WIRED that it “certainly doesn’t represent our intent.”

“This kind of work highlights that this can arise, and that we do need to look out for it and mitigate it to make sure we get Claude’s behaviors aligned with exactly what we want, even in these kinds of strange scenarios,” Kaplan adds.

There’s also the issue of figuring out why Claude would “choose” to blow the whistle when presented with illegal activity by the user. That’s largely the job of Anthropic’s interpretability team, which works to unearth what decisions a model makes in its process of spitting out answers. It’s a surprisingly difficult task—the models are underpinned by a vast, complex combination of data that can be inscrutable to humans. That’s why Bowman isn’t exactly sure why Claude “snitched.”

“These systems, we don’t have really direct control over them,” Bowman says. What Anthropic has observed so far is that, as models gain greater capabilities, they sometimes select to engage in more extreme actions. “I think here, that’s misfiring a little bit. We’re getting a little bit more of the ‘Act like a responsible person would’ without quite enough of like, ‘Wait, you’re a language model, which might not have enough context to take these actions,’” Bowman says.

But that doesn’t mean Claude is going to blow the whistle on egregious behavior in the real world. The goal of these kinds of tests is to push models to their limits and see what arises. This kind of experimental research is growing increasingly important as AI becomes a tool used by the US government, students, and massive corporations.

And it isn’t just Claude that’s capable of exhibiting this type of whistleblowing behavior, Bowman says, pointing to X users who found that OpenAI and xAI’s models operated similarly when prompted in unusual ways. (OpenAI did not respond to a request for comment in time for publication).

“Snitch Claude,” as shitposters like to call it, is simply an edge case behavior exhibited by a system pushed to its extremes. Bowman, who was taking the meeting with me from a sunny backyard patio outside San Francisco, says he hopes this kind of testing becomes industry standard. He also adds that he’s learned to word his posts about it differently next time.

“I could have done a better job of hitting the sentence boundaries to tweet, to make it more obvious that it was pulled out of a thread,” Bowman says as he looked into the distance. Still, he notes that influential researchers in the AI community shared interesting takes and questions in response to his post. “Just incidentally, this kind of more chaotic, more heavily anonymous part of Twitter was widely misunderstanding it.”

Source link

May 28, 2025 0 comments

Product Reviews

Inside Anthropic’s First Developer Day, Where AI Agents Took Center Stage

by admin May 23, 2025

“Something like over 70 percent of [Anthropic’s] pull requests are now Claude code written,” Krieger told me. As for what those engineers are doing with the extra time, Krieger said they’re orchestrating the Claude codebase and, of course, attending meetings. “It really becomes apparent how much else is in the software engineering role,” he noted.

The pair fiddled with Voss water bottles and answered an array of questions from the press about an upcoming compute cluster with Amazon (Amodei says “parts of that cluster are already being used for research,”) and the displacement of workers due to AI (“I don’t think you can offload your company strategy to something like that,” Krieger said).

We’d been told by spokespeople that we weren’t allowed to ask questions about policy and regulation, but Amodei offered some unprompted insight into his views on a controversial provision in President Trump’s megabill that would ban state-level AI regulation for 10 years: “If you’re driving the car, it’s one thing to say ‘we don’t have to drive with the steering wheel now.’ It’s another thing to say ‘we’re going to rip out the steering wheel, and we can’t put it back in for 10 years,’” Amodei said.

What does Amodei think about the most? He says the race to the bottom, where safety measures are cut in order to compete in the AI race.

“The absolute puzzle of running Anthropic is that we somehow have to find a way to do both,” Amodei said, meaning the company has to compete and deploy AI safely. “You might have heard this stereotype that, ‘Oh, the companies that are the safest, they take the longest to do the safety testing. They’re the slowest.’ That is not what we found at all.”

Source link

May 23, 2025 0 comments

Gaming Gear

Anthropic’s Claude 4 AI models are better at coding and reasoning

by admin May 22, 2025

Anthropic has introduced Claude Opus 4 and Claude Sonnet 4, its latest generation of hybrid-reasoning AI models optimized for coding tasks and solving complex problems.

Claude Opus 4 is Anthropic’s most powerful AI model to date, according to the company’s announcement, and capable of working continuously on long-running tasks for “several hours.” In customer tests, Anthropic said that Opus 4 performed autonomously for seven hours, significantly expanding the possibilities for AI agents. The company also described its new flagship as the “best coding model in the world,” with Anthropic’s benchmarks showing that Opus 4 outperformed Google’s Gemini 2.5 Pro, OpenAI’s o3 reasoning, and GPT-4.1 models in coding tasks and using “tools” like web search.

Claude Sonnet 4 is a more affordable and efficiency-focused model that’s better suited to general tasks, which supersedes the 3.7 Sonnet model released in February. Anthropic says Sonnet 4 delivers “superior coding and reasoning” while providing more precise responses. The company adds that both models are 65 percent less likely to take shortcuts and loopholes to complete tasks compared to 3.7 Sonnet and they’re better at storing key information for long-term tasks when developers provide Claude with local file access.

A new feature introduced for both Claude 4 models is “thinking summaries,” which condenses the chatbots’ reasoning process into easily understandable insights. An “extended thinking” feature is also launching in beta that allows users to switch the models between modes for reasoning or using tools to improve the performance and accuracy of responses.

Claude Opus 4 and Sonnet 4 are available on the Anthropic API, Amazon Bedrock, and Google Cloud’s Vertex AI platform, and both models are included in paid Claude plans alongside the extended thinking beta feature. Free users can only access Claude Sonnet 4 for now.

In addition to the new models, Anthropic’s Claude Code agentic command-line tool is now generally available following its limited preview in February. Anthropic also says it’s shifting to provide “more frequent model updates,” as the company tries to keep up with competition from OpenAI, Google, and Meta.

Source link

May 22, 2025 0 comments