In partnership with

The AI Field

Welcome back to The AI Field. Anthropic just told Claude users to stop writing "think hard" in their prompts, along with a short list of other habits that now slow its newest model down. Nvidia's biggest investors asked the board not to rush Jensen Huang's replacement. And Anthropic heads into the week it plans to go public, with its CEO calling compute its "deepest business dilemma."

Inside: three stories, five new AI tools, three quick hits, the most-viewed AI posts on X, and a five-minute way to get an AI briefing that never drops an open to-do.

In today's AI Field:

  • Anthropic's new prompting rules for Claude Opus 5.5

  • Nvidia investors ask the board not to rush Huang's replacement

  • Anthropic heads into IPO week with a warning about compute

Read time: 5 minutes

1,000+ Proven ChatGPT Prompts That Help You Work 10X Faster

ChatGPT is insanely powerful.

But most people waste 90% of its potential by using it like Google.

These 1,000+ proven ChatGPT prompts fix that and help you work 10X faster.

Sign up for Superhuman AI and get:

  • 1,000+ ready-to-use prompts to solve problems in minutes instead of hours—tested & used by 1M+ professionals

  • Superhuman AI newsletter (3 min daily) so you keep learning new AI tools & tutorials to stay ahead in your career—the prompts are just the beginning

LATEST DEVELOPMENTS

ANTHROPIC · CLAUDE

Anthropic's seven prompting habits for Claude Opus 5.5

Source: Anthropic

The AI Field: Anthropic published a guide this week for Claude Opus 5.5, now the default model on the Pro, Max, Ultra and Team plans. Its message is simple: every new model is a reason to retest how you prompt, and to drop habits that were built for older models.

The details:

  • Delete "think hard": Thinking is always on in Opus 5.5, and it decides how much to think from your request and the effort level. In Anthropic's testing, removing those lines made replies start sooner, with no clear drop in quality.

  • Start at Medium effort: Medium is the default, and Anthropic says Opus 5.5 on Medium matched or beat Opus 5 on High in its coding and knowledge-work tests. Move up to High only when Medium stalls. Click the model name next to the send button and choose Effort.

  • Say what done looks like: Put the goal, the finished output and when Claude should stop and ask in your first message, instead of feeding instructions one at a time. Ask for the finished spreadsheet or document, not an outline.

  • Attach, don't retype: Share the chart or screenshot itself instead of copying its numbers. For design work, name the styles you don't want, like a cream background or pill-shaped buttons, instead of saying "avoid a generic look."

Why it matters: Prompt tricks from a year or two ago now mostly cost time. The fastest test is Anthropic's own: take a job you would normally split across several messages, give Opus 5.5 the files and one description of the finished result, leave it on Medium, and compare the output with what you usually get.

The Agent Field

TRY THIS TODAY · 5 MINUTES

Get a daily AI briefing that never drops an open to-do

A student's AI secretary in Claude Duet reads Gmail, Google Calendar, Canvas and a Google Sheets tracker, and sends a morning and an evening briefing. One evening it left out a quiz because it had already been in the morning digest. The fix was one rule: old news can still need action.

1. Pick what your agent should read: your email, your calendar and any course or project site.

2. Paste this prompt into your agent:

"Each morning and evening, send me a briefing from my email, calendar and course site. Always list every open deadline, quiz and task due in the next 7 days, even if you mentioned it in an earlier briefing. Mark new items as new. Only drop an item when it's done or I tell you to. Keep it under 15 lines."

3. Run it for a week next to your own calendar before you rely on it. The same setup ran in Grok Bot and a ChatGPT Dot before moving to Claude Duet, so it works in any agent that can read your email and calendar on a schedule.

TRENDING AI TOOLS

Five new tools to try today. Click a name to open the tool's own site.

⏱️ Burnstate: Keep hitting your Claude Code or Codex limit mid-task? Burnstate shows your AI coding limits from Claude, Codex, Cursor and Vercel in your Mac's menu bar and warns you hours before you'd run out. From $3.99 a month, or $49.99 once.

🎙️ FreeFlow: Hold the Option key, talk, and your words appear at the cursor in any app, with the filler words taken out. Transcription runs on your Mac, so nothing is uploaded. Free for 14 days, then $99 once. Needs Apple Silicon and macOS 14 or later.

📈 GAtrends: Pick the keywords that matter to your business and GAtrends sends you a Telegram message the moment one spikes across news, search and online communities. Paid plans let you ask Claude or ChatGPT what is rising and why. Free for 3 keywords.

🗒️ VB Notes: Your notes stay as plain markdown files in a folder you choose, and search finds them by meaning, even when you can't remember a single word. The optional AI chat runs on your own API key. Free, no account, on Mac, Windows, Linux, iPhone and Android.

🔁 Replicate: Built your site in Webflow, Framer, Squarespace or Wix? Replicate turns it into a code project on your GitHub, checked page by page against the original, ready to edit in Claude Code, Codex or Cursor. Free.

👉 Browse 130+ more free and paid AI tools in our AI Tools directory

Reach over 8 million engaged tech professionals

The tech decision makers you're trying to reach read TLDR every morning: 8 million tech professionals across 14 curated newsletters, with a 41%+ average open rate. Paid social can buy reach, not attention. Plaid drove 20.1x ROI from its last TLDR campaign.

NVIDIA · LEADERSHIP

X post: Altimeter Capital CEO Brad Gerstner shares the letter. Source: CNBC

The AI Field: Two major Nvidia shareholders, Altimeter Capital and Atreides Management, sent Nvidia's board a letter on Friday asking for "a slow, measured, and deliberate process" in choosing its next leader. They want every option considered, including "the potential return of Jensen Huang," CNBC reports.

The details:

  • Who signed: Altimeter CEO Brad Gerstner posted the letter on X, co-signed by Atreides' Gavin Baker. Altimeter held about $1.9 billion of Nvidia stock as of June 30.

  • Not a comeback demand: The letter says interim CEO Colette Kress, Nvidia's CFO for 13 years, will lead the company "capably," and it doesn't call for Huang's immediate return.

  • The pressure: An online petition called "Stand with Jensen" claims more than 100,000 signatures, including more than 5,000 current Nvidia employees. Huang resigned on Nov. 12 amid an independent board investigation.

  • Washington: Commerce Secretary Howard Lutnick has agreed to testify before the Senate Banking Committee on Dec. 10 about keeping advanced AI chips out of China, CNBC reports, and Huang has also been asked to testify.

Why it matters: Nvidia supplies the chips behind much of the AI industry, so a drawn-out leadership fight matters well beyond its share price. The letter asks the board to wait until "all the facts are known." The Dec. 10 hearing is the next date to watch.

ANTHROPIC · IPO

Anthropic IPO site artwork showing three illustrated trees

Image: Anthropic. Source: The Decoder, Bloomberg

The AI Field: Anthropic expects to set its IPO price on Wednesday, Nov. 18, and start trading on Nasdaq as ANT on Thursday, Nov. 19 (US Eastern time), with an estimated range of $70 to $78 a share. At a JPMorgan event with Jamie Dimon on Friday, CEO Dario Amodei was blunt about the company's biggest risk.

The details:

  • The dilemma: "Compute is the deepest business dilemma that the company faces," Amodei said. Chips have to be bought one to two years ahead, before anyone knows if revenue will cover them, and "you can financially overextend yourself."

  • The hedge: He said Anthropic buys relatively little compute upfront and turns to the spot market if business runs hot. More cash means a wider margin for error, which he called the biggest reason to go public.

  • The size: Bloomberg reports Anthropic plans to raise up to $64.4 billion in the IPO plus $26 billion in a private placement, which together would make it the largest IPO ever.

Why it matters: Retail investors in nine European countries, including Sweden, Norway, Denmark and Finland, plus Switzerland, will pay the same IPO price as everyone else. Amodei's own summary of the risk is worth keeping in mind: "There's a huge uncertainty in what's going to happen next year." This isn't investment advice, so read the prospectus before you decide anything.

QUICK HITS

🧩 Meta's Muse moves into Chrome: A "Muse from Meta" extension opens the agent in Chrome's side panel to fill out forms, compare prices and follow up on refunds, asking before it acts and keeping a full activity log. The Chrome Web Store lists 90,000 users.

🛒 Amazon wants to index everything for sale online: An internal document seen by Reuters describes a "World Product Graph" covering products sold anywhere, even ones Amazon doesn't sell, so shoppers and AI agents end up on Amazon. Amazon says the brainstorming document doesn't necessarily reflect its current plans.

🧮 Gemini 4 Carbon tops a math leaderboard: ETH Zurich's MathArena added Gemini 4 Carbon and Claude Fable 5.2 on Nov. 13. The Decoder reports Carbon took first place from GPT-6.1 Sol, with Fable 5.2 second, but cost about 15 times as much to run at the settings tested.

TRENDING ON X

That's it for today!

Was this edition worth your time? Leave a rating to help me make The AI Field more useful for you.

Login or Subscribe to participate

See you soon,

Olle Hellman | Founder of The AI Field