In partnership with
200+ Claude Prompts Top Professionals Actually Use at Work
Claude can be your analyst, editor, and strategist.
But most professionals are using it to fix grammar.
These 200+ Claude prompts take it from grammar tool to your most powerful AI work assistant.
Sign up for Superhuman AI and get:
200+ ready-to-use Claude prompts to get real work done in minutes — researched, tested, and used by professionals at Google, Microsoft, and NASA
Superhuman AI newsletter (4 min daily) so you keep learning new AI tools and skills to stay ahead in your career — the prompts are just the beginning
TODAY IN AI
The AI Edge
3 things that happened while you were busy
1. Grok 4.5 arrives, gunning for the frontier at a discount.
SpaceXAI released its first model since going public, built for coding, agents, and knowledge work, and trained alongside Cursor after the reported $60B acquisition. Musk pitches it as an "Opus-class" model, and the pricing is the real headline: $2 per million input tokens and $6 output, against $5 and $25 for Claude Opus 4.8. One honest caveat: the benchmark charts are vendor-published, so test it on your own work before switching anything important.
2. Fable 5's extended deadline lands today.
Anthropic quietly extended paid-plan access to Claude Fable 5 through July 12, pushing back the usage-credit switchover it announced last week. That grace period ends today, and the announcement did not say what happens next. Translation: if you have a Fable-sized task waiting, run it this evening rather than tomorrow morning.
3. Google Photos will now repaint your videos.
A new Video Remix feature powered by Gemini Omni turns ordinary clips into stylized ones in seconds: cinematic relighting, background swaps, and effects like watercolor, sketchbook, and oil painting, all from templates in the Create tab. It is rolling out now to Google AI Plus, Pro, and Ultra subscribers in select countries.
FROM THE FRONTIER
The AI Edge
The frontier AI price war is on, and you are the winner.
The new bar. Grok 4.5's launch is less about benchmarks and more about a number: frontier-adjacent capability at $2 in and $6 out. A year ago that class of model cost five to ten times more. SpaceXAI can price this way partly because it owns its compute, the same Colossus capacity it leases to Anthropic and Google, which sets up an interesting conflict as its own needs grow.
The pile-on. Everyone is cutting at once. GPT-5.6 Terra matches GPT-5.5 at half the price, and Luna runs $1 in and $6 out. Claude Sonnet 5 launched at $2 introductory pricing through August. The premium tier still exists, but the gap between "good enough" and "the best" has never been cheaper to ignore.
The twist. The cheapest supply may be about to tighten. China is reportedly weighing foreign-use limits on top models from Alibaba, ByteDance, and Zhipu, the same models that grew to serve roughly 45% of OpenRouter traffic on price alone. If those limits land, Western budget tiers become the only game in town, and today's prices might be the floor rather than the ceiling.
The takeaway. Model pricing now changes monthly, but most people set their AI subscriptions once and forget them. The winning habit is routing: match each task to the cheapest tier that does it well, reserve premium models for the work that genuinely needs them, and rerun the math every quarter. The prompt below builds your routing rules in one conversation.
IN THE KNOW
What people are actually watching and sharing
Apple's fine print. Apple Home's new AI features, including searchable summaries of your security camera footage, require a 2TB iCloud+ plan at $10 a month or the $37.95 Apple One Premier bundle. The smart home just got a subscription tax.
Musk's victory lap. Musk spent launch week posting benchmark wins, with one claim-heavy post passing 465,000 views, and confirmed a 1M-plus token context window is coming to Grok. Vendor scoreboards aside, the pace of his shipping cadence is the real flex.
Lovable's double. Vibe-coding startup Lovable is reportedly raising $300M at a $13.2B valuation, double its December number, on the back of $500M in annualized revenue. Build-an-app-by-chatting is now a real business category.
Watermark wins one. Google's SynthID detector helped debunk a viral fake hospital photo of Senator Mitch McConnell by spotting the invisible AI watermark. A rare public win for provenance tech in the deepfake era.
PROMPT STATION
Build your personal model-routing cheat sheet
With Grok 4.5 at $2, Luna at $1, and premium models at $5-plus, paying flagship prices for routine tasks is the new leaving-the-lights-on. Paste this into Claude or ChatGPT and walk away with a one-sentence routing rule that fits how you actually work.
You are an AI operations advisor who optimizes for value, not prestige. Here are the AI tasks I do in a typical week: [TASK 1], [TASK 2], [TASK 3], [TASK 4], [TASK 5]. Current model pricing comes in three rough tiers: budget (around $1 per million input tokens), mid-tier ($2 to $3), and premium ($5 and up). For each task, tell me the cheapest tier that would handle it with no noticeable quality loss, and explain your reasoning in one line. Flag any task where budget and premium output would be indistinguishable to me. Then write my personal routing rule as a single memorable sentence. Finish by naming the one task on my list where paying for premium genuinely pays for itself, and why.
Fill the task slots with real work like "summarize long PDFs", "draft client emails", "debug Python scripts", "brainstorm content ideas", "clean spreadsheet data". Advanced tip: add "assume my time costs [YOUR RATE] per hour" and it will tell you when a pricier model that gets it right the first time beats a cheap one you have to babysit.