If you only check one highlight this week: AI hackathons. Run them. Participate in them. At least quarterly.
Until things stabilize.
But recently - and I think most people may miss that - they've drastically accelerated. I'm sharing this given my vantage point in the AI programming space.
-François
The impact of AI hackathons + a playbook
I already knew team hackathons create an incredible impact. Especially with AI, since it's so easy to use.
Two weeks ago, Amplitude’s CMO invited me to deliver a keynote at their Marketing and Ops teams' AI Day and to serve as a judge at their AI hackathon.
That was a powerful reminder of how impressive they are in driving AI adoption, excitement, innovation, and team collaboration.
I highly, highly recommend you do this at least once a month with your team.
Impact
Proficiency and excitement for using AI and building agents increased drastically in just a few hours. Many realized how easy it is. The teams had fun. It felt like a real unlock for many.
C-suite execs committed to funding projects and new agents on the spot.
Hackathon teams build hacks that will eventually deliver significant value to customers and the business. Many of them have the potential to replace expensive SaaS.
Projects were creative, tackled significant business needs and customer problems.
The playbook
Here's what they did right to inspire their team and unleash their builder creativity:
1. Created a collective space: they blocked a whole day at their company kick-off.
2. Before the AI day: Set up the tools and context first. Amplitude's IT and ops team had done the hard work beforehand. They selected, approved, and set up all the right tools + integrations in advance, laying the groundwork for each team to build with the right tools and context: a central brain connected to their customer and business data. That deep context made the AI workflows & agents easy to build and powerful.
3. To kick off the day: Execs set context and expectations. Ensured the team understood that AI proficiency is a necessity, not a luxury. They brought me on next to reinforce that and show how easy and powerful it can be + share my story and workflows + some external inspiration.
4. Before the hackathon:
Team members showcased a few of their workflows as inspiration. They showed how they work and how they built them.
Had IT and Ops pros demo and teach two AI tools they had built or set up. especially the agent builder interface.
5. For/during the hackathon:
Mixed AI-comfortable people with those who aren't. Each brought their respective expertise.
Each team had a director or VP as exec sponsor/leader
The IT and Ops teams were distributed across teams and could be tapped for support
6. Selecting winners
The top teams pitched their projects to a panel of judges in front of the whole team. The projects impressed and inspired.
The jury assessed projects based on potential business impact, relevance of the problem, and likelihood to scale. We asked questions. We explained the rationales for our selections.
7. C-level involvement. The CEO and President attended. They saw what teams built in two hours. They were impressed and committed to funding projects on the spot.
8. Committed to repeating it, on the spot. The hunger to do more was palpable.
How I AI: an update on my go-to LLMs
ChatGPT used to be my go-to.
That's changed a lot.
Here are my top three now (only covering models below, not point solutions):
🏅 Claude
You've probably guessed by now, but the Claude desktop app has become my go-to partner for most work tasks.
I am constantly switching between Chat, Cowork (mostly to work on local files) and Code mode. With Opus 4.5.
Claude is where I store most of my work context, start most of my work, my strategies, content creation, run analysis, and code my apps. I have many Projects (at least one per client and major activity or frequent workflow that needs stored context) and skills there.
I pay for the $100/month Max tier. Given the time savings and value it provides, it feels like a total bargain vs. paying an analyst.
🥈 Google/Gemini
Gemini 3, Veo3 and Nano Banana (image gen)are really good. So is NotebookLM. I typically use Gemini because most of my files are in Google Workspace and I use Chrome. The integration with Workspace provides great convenience for quickly summarizing a document, creating content from it, conducting analyses or critiques, and generating images. I love opening a contract in my Google Drive and asking Gemini to give me an overview and recommendations on what I should negotiate and how, and write the follow-up/negotiation email for me.
I wish they had a desktop version… I pay for the $20/month. Since that also includes increased storage, it’s a steal.
🥉 ChatGPT
Downgraded from number one not so long ago for a variety of reasons:
the two listed above are really good. I've now migrated my go-to custom GPTs into Claude skills
I use ChatGPT mostly for convenience (E.g., “transcribe this screenshot”) and personal stuff - and to avoid “polluting”/crowding the work context that I increasingly have in Claude and Gemini
a strong reminder that brand matters. Anthropic insists on security and is very thoughtful about sharing Claude’s constitution. Google has always handled our data with care. But OpenAI is the brash brand that brought us Sora, the AI-video slop social app the world doesn’t need, when it claims to be building AGI (#endofrant)
the advanced voice mode of ChatGPT remains the best, though
I pay for the $20/month tier, yet I've realized it often feels lazy. Maybe because it needs to generate all these Sora slop videos?
My selection of tips, news and workflows
📚 40 proven GTM go-to-market AI plays
Kyle Poyar and Maja Voje interviewed 30 GTM experts who are seeing real gains from AI, and compiled 40 of their best AI x GTM plays.
Kyle details them in his latest report and even shares the workflows and/or prompts.
🐝🐝🐝 Agent swarms: understanding how we will work in the near future
AI-native developers already work like this by orchestrating agents.
But this launch video for Kimi K2.5 Office and Kimi agent swarms shows us how we'll soon be working.
Does that scare or excite you?
📚 How to use Claude Code for non-technical folks
I enjoyed this tutorial. It includes a very impressive use case of building competitive campaigns in no time.🤯
The instructor uses Claude Code in VS Code, but you really don't need to. Use Claude’s desktop app instead. There, you can access Chat, Cowork, and Claude Code at the same time.
It has become my go-to tool for work.
📨 Outbound Inspiration
Stevie Case, CRO at Vanta, shared a cool personal direct mailer and gift that she received, and it definitely got her attention.
🖊️ Handwritten-ish inspiration
Want to grab the attention of your readers on LinkedIn or whatever social feed? Charlie Hills shares his workflow with Gemini 3 Pro to create handwritten-looking images and post them on LinkedIn.
🗣️Prompting tip
Do not start with a prompt; tell the AI your goal first.
Ask the AI to interview you with 5-10 questions to get the context and instructions it needs (see the final words at the very bottom of this edition).
👨💻 This is what working with AI should feel like now
I completely agree with Francis, who articulates how working with AI has changed over the last few months as he creates software.
That shift is coming to non-developers. Already here, actually.
This new way of working is similar to a marketer writing a very good strategic brief to an agency with context, goals, target audience, constraints, etc., and then redirecting with every round, until the outcome is world-class.
Final Words
In agent-driven systems, the builders who succeed aren’t the ones who prompt most cleverly. They’re the ones who can make intent explicit and constraints clear.
Thanks for sharing these highlights with busy marketing execs around you.🙏
Someone forwarded you this email? You can subscribe here.
François | LinkedIn
I'm a CMO, advisor, and "CMO Wingman". Yes, that's a thing :-). Ask my clients: in this AI era, CMOs need a strategic proactive advisor more than ever. I’m former CMO at Twilio, Augment Code, Apollo GraphQL, Decibel, Udacity and Head of Marketing for LinkedIn Talent Solutions.
