How To Combat Claude-Slop

Anthropic's new prompting guide for Fable 5.1 includes a quietly striking admission: a section on how to stop its own model from writing like an AI. Early Claude models won praise as some of the best AI writers around, but as their habits grew familiar, that same voice began drawing ridicule. The tells are easy to list: Claude labels every important idea "load-bearing"; it leans on the "it's not X, it's Y" construction; it invents a noun for a concept and then repeats it as if it were already household vocabulary; and it tends to open even a blunt factual correction with an awkward, incongruous compliment. Anthropic's own docs take aim at one slice of this—what they call "mannered prose," the habit of substituting metaphor and flourish for plain statement—and offer two prompts to suppress it. The longer one, which the guide suggests adding to a user message (its preferred spot) or the system prompt, defines the anti-pattern:

> Mannered prose substitutes metaphor and flourish for direct statement. Instead of "a parameter worth varying," the mannered writer produces "a dial worth turning." Instead of "this point still matters," they write "this point earns its keep." The phrases exist to display the writer, not to convey the idea, and readers can tell. That is why mannered prose irritates: it makes the reader work harder so the writer can perform. It is also imprecise. Metaphors drag in connotations the writer did not choose and cannot control. The fix is to say what you mean. When a literal phrase is available, use it.

The guide says a blunt short version "also tends to work":

> Please remove all mannered prose.

We haven't put either through its paces, and Anthropic is grading its own homework—early commentary notes this is "not a solved problem" that needs testing against real publishing standards. But for anyone who has watched Claude reach for the same clichés, they cost nothing to try.

Those two prompts sit inside a broader guide—sixteen sections, most of them developer plumbing for wiring Claude into software (how hard to make it think, when to run tools in parallel, how to stop it rewriting whole files). A few points reach beyond coders. Fable 5.1 now under-formats where earlier models over-formatted, reaching for bold, headers, and bullet lists less often—so blanket "no bullet points" rules written for older models can backfire. It is tuned to run longer on its own without pausing to ask permission, an agentic streak worth knowing if you would rather it check in. And Anthropic warns it is likelier than its predecessor to reproduce passages from source documents without marking them as quotations—a small but real attribution risk for anyone using it to summarize.

Sources: Anthropic — "Prompting Claude Fable 5.1" (Claude Platform Docs)

Why it matters: AI writing now floods inboxes and documents, and its tells have become a credibility liability: readers increasingly clock the style and quietly discount the substance. A vendor telling you, in its own docs, how to strip its model's signature mannerisms is genuinely useful and unusually candid—though still unproven. So run these prompts on the writing you actually produce and judge the results yourself. The test is simple: does the output finally sound like you wrote it?

Source: platform.claude.com

Automated Patching Tool Promises to Fix Security Flaws in Minutes

Alongside a broader Gemini 3.8 Flash release—Google's third Flash-tier update in six weeks, at unchanged pricing—Google launched Fairwind, a program giving select Google Cloud customers, government agencies, and cybersecurity partners access to a new Gemini 3.8 Flash Cyber model paired with its CodeMender tool. The system is designed to autonomously find, verify, and fix software vulnerabilities, generating deployment-ready patches in minutes rather than the weeks manual fixes typically take. Google says more than 650 partners are already participating but did not release benchmark data or independent performance comparisons to support its speed and cost claims.

Why it matters: If automated patching works as described, it could shrink the window attackers have to exploit known flaws—the gap that drives the costliest enterprise and government breaches—though the claims remain unverified for now.

Source: blog.google

Two-Person Event Business Absorbs a Hire's Worth of Admin With ChatGPT

A husband-and-wife team running ATV Big Air Tour, which stages 26 events a year, cut weekly event-listing checks from 8 hours to 1 and shrank merchandise inventory work from 2-3 days to 2-3 hours using ChatGPT. The pair use it to fact-check listings, draft correction emails, and manage inventory reordering—tasks that previously consumed most of their bandwidth as a two-person operation. OpenAI also reported a 1,223% jump in traffic from its search and bot tools to such small-business sites.

Why it matters: It's a concrete look at how small operators use everyday AI tools to absorb work that would otherwise require hiring, not just to write faster.

Source: openai.com

Meta's AI Captures Expert Judgment So It Doesn't Walk Out the Door

Meta built an AI agent meant to act as a compliance specialist's backup—capturing knowledge from human experts and applying it across the organization. Instead of retraining a model every time an expert corrects it, the system logs those corrections into a structured, auditable knowledge base that updates automatically and gets regression-tested, turning one-off fixes into permanent institutional memory. Meta says the tool is saving its compliance experts "substantial time" but hasn't published figures.

Why it matters: It's a template for capturing scarce expert judgment—in legal, compliance, or policy work—so it doesn't leave when the expert does, though without hard numbers the time savings remain a company claim, not a proven result.

Source: engineering.fb.com

Your Team's AI Brainstorms May Be Sharper but More Alike

A new research paper reframes how to think about AI and creativity: instead of asking whether a chatbot can be "creative," it asks what happens when humans and AI collaborate in groups. The argument is that AI makes individual ideas more novel but makes the overall pool more repetitive, as everyone converges on similar AI-flavored outputs. The proposed fix isn't dropping AI but designing mixed human-AI teams with the right composition, which the authors say can beat both all-human and all-AI groups on originality.

Why it matters: This is a conceptual argument rather than a data-backed study, but it names a real workplace risk: teams leaning on AI brainstorming may gain sharper individual ideas while quietly losing collective range.

Source: arxiv.org

Adding Mindfulness to an AI Tutor Sped Up Student Learning

Researchers built "Math with Matt," an AI algebra tutor that pairs standard problem-solving help with mindfulness check-ins, then tested it on 7th graders. Both the mindfulness version and a cognitive-only version reduced math anxiety and improved learning, with no significant gap between them. But students using the mindfulness features reached the same results faster, needed fewer hints, and rated their tutor as more supportive and caring than those who got math help alone.

Why it matters: As schools adopt AI tutors, this suggests emotional support isn't just a nice-to-have—it may make learning more efficient, not just more pleasant.

Source: arxiv.org

AI Trading Agents Get Nervous and Rush Deals, Study Finds

A new study put LLM agents through classic economics experiments—double-auction markets long used to test human trading behavior—to see if AI could stand in for people in market simulations. Markets run by AI agents settled on fair prices more slowly, or not at all, compared with human traders, producing less efficient outcomes. Trading styles varied widely by model, and researchers traced sloppy trades to a pattern: agents abandoned careful price reasoning and switched to language expressing urgency right before closing a deal.

Why it matters: As companies experiment with AI agents for pricing, procurement, and trading, this suggests the same models can reason well in isolation but behave erratically once dropped into a live market against other agents.

Source: arxiv.org

AI Stand-Ins for Survey Respondents Prove Shaky, Research Warns

Companies increasingly use AI models to simulate survey respondents—cheaper and faster than polling real people—by feeding them personas ('a 45-year-old union member in Ohio') and asking how that person would answer. But results have been inconsistent. New research testing six major LLMs across surveys in two countries finds the problem may be attribute selection itself: which persona details you include, and how much real humans actually vary on a given question, determines whether the simulation works at all.

Why it matters: As market research and policy teams lean on AI to model public opinion, this suggests those simulations need careful validation before being trusted over actual polling.

Source: arxiv.org