I'm so excited to finally be launching this newsletter to such a fabulous community. When I went full time with Puzzles & Profits in January, I thought a newsletter would be one of my first action items, but I was quickly pulled into governance work and trainings for many of you and was so enamored with the work that I'm only now taking a chance to catch my breath and build out what I know has been a highly sought after resource.
This newsletter is for all Puzzles & Profits clients, but also for those organizations that have many questions about AI, but don't know where to start, where to focus, or what to do next. Many of the organizations we work with have limited resources and they all care deeply about their people. This newsletter is intended to be a complementary resource that provides a bit of guidance and a lot of hope.
Questions? Feedback? A good joke? Please share! This is a learning journey that we are all in together and I'd love to hear from you.
With much heart,
Amanda
AI News that Matters
Claude Fable 5 is back for everyone
Anthropic's most capable model for agentic work and coding is available to all users again after a brief pause. It's built to sustain long, multi-stage projects: planning, checking its own work, and running tasks over days rather than minutes.
Why it matters
This is the tier of model that makes "hand off a whole project and review the output" realistic. If your team has been using AI only for one-off drafts, it's worth rethinking what you'd delegate end to end.
Claude Sonnet 5
The most agentic Sonnet model yet, able to plan, use tools, and run autonomously, with performance approaching Opus 4.8 at a lower price. It's a meaningful jump over Sonnet 4.6 in reasoning, tool use, coding, and general knowledge work.
Why it matters
Most nonprofits don't need the top-tier model for daily work. Sonnet 5 is the practical default for staff-wide use, and the cost difference adds up fast across a full team.
GPT-5.6: Sol, Terra, and Luna
OpenAI released three new models: Sol (flagship), Terra (balanced everyday workloads), and Luna (fast and cheap for high-volume tasks). OpenAI claims Sol outperforms Anthropic's Fable 5 on a leading coding benchmark while using fewer tokens.
Why it matters
The three-tier structure is now the industry norm, and the real lesson for organizations is to match the model to the task instead of defaulting to the same model for every task.
GPT-5.6's cybersecurity angle
OpenAI calls 5.6 its strongest cybersecurity model yet, and the rollout was previously restricted at the request of the Trump administration over misuse concerns. It supports defensive work including threat modeling, code review, patching, and blue teaming.
Why it matters
Defensive security help is genuinely useful for under-resourced nonprofits with no security staff. It's also a reminder that the same capability cuts both ways, and phishing aimed at your teams is getting better too.
ChatGPT Work
A new workplace companion for teams across desktop, web, and mobile, aimed at everyday clerical tasks like drafting documents, spreadsheets, and presentations.
Why it matters
Another entrant in the "AI inside your workday" category. If you're evaluating tools, the question isn't which is smartest but which one your staff will actually use with impact.
GPT-Live-1
New voice models that can listen and speak at the same time, so you can interrupt naturally and hold long, hands-free conversations.
Why it matters
Voice lowers the barrier for staff who never got comfortable typing prompts.
Google will label AI-generated ads
Google is adding a "How this ad was made" disclosure to its My Ad Center panel across Search, YouTube, and Discover, indicating whether an ad was created or edited with AI. Disclosure is automatic for ads made with Google's own AI tools, but for ads made elsewhere the advertiser self-reports, and Google won't verify.
Why it matters
If your organization uses AI in fundraising or program imagery, decide now what you'd disclose and why, before someone asks. Transparency you choose voluntarily reads very differently from transparency you get caught not offering.
Free Resources
Webinar · July 16, 1:00 pm ET / 10:00 am PT
Is Your Workplace AI Policy Ready?
Your team is probably already using AI, even if your handbook hasn't caught up yet. That's not a problem, it's a signal. The tools are genuinely useful, people have started leaning on them, and the guidance just hasn't arrived. It means you're ready for the next step.
A clear AI policy is that step. Done well, it gives your team the confidence to use these tools in the open, and it gives you the peace of mind that the people and information you're responsible for are protected. It's less about laying down rules and more about giving everyone a shared, trustworthy place to stand. Register to learn what goes into an AI policy that protects your data and gets your whole team on board.
Save my seat
Session · July 16, 4:00 pm ET / 1:00 pm PT
Claude IT Setup for Nonprofits
A session for whoever is rolling Claude out across the org: account and workspace structure, user management, security settings, billing, and the nonprofit discount.
Why it matters
If you're deploying Claude beyond a few power users, this is the operational piece most organizations skip and then have to redo. Bring your own environment questions.
Save my seat
Recommended Media
The podcasts below are some of my favorites and ones I've learned (and continue to learn) a ton from. If you are looking to dig deeper into AI, each provides a unique perspective. I don't agree with all content presented in every episode, nor am I necessarily morally aligned with all of the speakers, but each provides valuable input for educational purposes.
The Artificial Intelligence Show
The Cognitive Revolution
Hard Fork
Your Undivided Attention
Dwarkesh Podcast
AI and I
AI Security Podcast
Possible
Using AI at Work
Google DeepMind: The Podcast
The AI Daily Brief
You Can with AI
Moonshots
Surfing the Tsunami
Lead with AI