Friday, April 17, 2026

mAIn

STREET

Today's throughline

In what is probably the biggest news today, Anthropic has released Claude Opus 4.7. According to the company's blog post, you can hand over your toughest coding work to it without supervision. It can handle complex, long-running tasks with rigor and consistency, pays precise attention to instructions, and devises ways to verify its own outputs before reporting back.

Below are some highlights and notes from early testing:

Instruction following. Opus 4.7 is substantially better at following instructions. Interestingly, this means that prompts written for earlier models can sometimes now produce unexpected results: where previous models interpreted instructions loosely or skipped parts entirely, Opus 4.7 takes the instructions literally. Users should re-tune their prompts and harnesses accordingly.

Improved multimodal support. Opus 4.7 has better vision for high-resolution images: it can accept images up to 2,576 pixels on the long edge (~3.75 megapixels), more than three times as many as prior Claude models. This opens up a wealth of multimodal uses that depend on fine visual detail: computer-use agents reading dense screenshots, data extractions from complex diagrams, and work that needs pixel-perfect references.

Real-world work. As well as its state-of-the-art score on the Finance Agent evaluation (see table above), our internal testing showed Opus 4.7 to be a more effective finance analyst than Opus 4.6, producing rigorous analyses and models, more professional presentations, and tighter integration across tasks. Opus 4.7 is also state-of-the-art on GDPval-AA, a third-party evaluation of economically valuable knowledge work across finance, legal, and other domains.

Memory. Opus 4.7 is better at using file system-based memory. It remembers important notes across long, multi-session work, and uses them to move on to new tasks that, as a result, need less up-front context.

You'll need a paid plan (at least $20 a month) to try it out. Word of warning, though: Anthropic is ruthless with its usage limits, so you may not get very far unless you're willing to Max out (from $100 monthly).

Prompt

Turn complaint logs into a one-week fix plan

Use this when you manage support, retail, or service operations and need to decide what to fix first before the next shift cycle.

I'm going to paste customer complaints, survey comments, and support tickets from the last two weeks. Group them into root causes, rank them by likely impact on revenue, retention, or reputation, and tell me which problems I should fix in 48 hours, 7 days, and 30 days. Then write a one-page shift-manager briefing with the top three issues, the coaching point for staff, and the process change we should test next week.

Prompt

Write a hiring debrief the team can defend

Use this after interviews when notes are messy, inconsistent, or drifting toward gut feel instead of evidence.

You are helping me run a fair hiring process. I'm going to paste the job description, interview scorecard, and notes from each interviewer. Build a structured debrief that summarizes evidence for strengths, risks, role-fit, and missing information. Flag any feedback that sounds biased, vague, or unsupported, then give me the three follow-up questions or reference-check questions that would most reduce uncertainty before we decide.

Prompt

Build a renewal rescue plan from account notes

Use this when you manage client accounts and need a sharp plan before a shaky renewal turns into a churn surprise.

I'm going to paste CRM notes, recent meeting notes, support issues, usage data, and the renewal date for one customer. Diagnose the top churn risks, tell me what evidence supports each one, and build a 14-day recovery plan with outreach timing, talking points, internal owners, and the one concession or offer that is most likely to help without giving away margin too early. End with a concise QBR agenda that gets the account back on stable ground.

Recall 2.0

Recall 2.0 turns saved articles, notes, PDFs, videos, and highlights into a personal AI knowledge base you can actually query later instead of forgetting where everything went.

For knowledge workers, the value is simple: it helps turn content hoarding into reusable context. If your work depends on remembering what you read, what you saved, and why it mattered, this is much more practical than yet another blank chatbot.

Source: Recall

Work & Productivity

Business & Strategy

Trust, Safety & Governance

Markets & Infrastructure

Consumers, Learning & Science

Law, Privacy & Governance

Health, Media & Society

Products, Pricing & Interfaces

Infrastructure & Industrial AI