Today’s Episode
90% of teams have adopted AI, yet 56% of CEOs say they saw no major financial benefit. Both metrics are accurate, and I’m pretty sure you fall somewhere between them.
With new AI tools flooding the market every week, it is safe to say that every team has its own AI setup by now. But are any of those setups in sync with each other?
I’ve been building this argument in stages. First, Carl Vellotti showed you how to build a personal OS, and Hannah layered it with building a team OS. Jiaona and Mikhail vouch for a full-blown company OS as well.
So, today you get a screen share.
My guests are the product team at Together AI that raised $800M at an $8.3B valuation and sells inference and fine-tuning to developers. Charles Zedlewski is their CPO who brought Necoline, Pavneet, and Hassan on the call to talk about how to build a shared context repo.
Their customers are already agents and their engineers are already agent-first. Their product team had no choice but to catch up.
Brought to you by
Arize - Ship AI agents and features faster with fewer regressions.
Get a full year of Arize, Bolt.new, Airtable, Speechify, Descript, Magic Patterns, Linear, Dovetail and Mobbin at bundle.aakashg.com.
10 Key takeaways
1. Individual productivity can move a company backwards - The team's starting question was not how to make each person faster. It was whether everyone generating unlimited code and content actually added up to progress. Charles called the failure mode flooding your coworkers' context windows, where everyone launches slop at each other.
2. The shared repo holds context and skills, not code - Markdown and YAML files covering customer intelligence, sandboxes, and the output of strategy meetings broken down by mission and milestone. Anything tied to a specific codebase stays out of it. The point is that a PM can read another team's context and draft a real proposal before taking up that PM's time.
3. Skills live closest to the work they touch - If a skill references code inside one team's repo, it stays colocated there. Everything else goes to a personal or shared repo. Test it on a branch, use it a few times, and only push to main once it proves repeatable. Niche ones never get pushed.
4. Shared context is a hierarchy, not a flat pool - The team abandoned the idea that everyone should carry everyone's context. Most people have no motivation to learn the depth of someone else's area. They want the one answer they came for. Some people live at the bottom of the hierarchy, most just traverse the top.
5. The PRD stopped being a gate - Historically it was the document everyone aligned on before building started. Together treats it as a trigger for ideation and problem solving instead. One to two pages, defining the customer problem, a few solution options, and a sample user journey. That is enough to argue about whether the thing is worth building.
6. A prototype replaces the bulk of the long document - A separate skill takes the one pager and produces a prompt for a design tool, and that visual is where the sharpest feedback shows up, from engineering and marketing alike.
7. Discovery collapsed from half a day to five minutes - The research agent pulls from the support platform, the project tracker, and internal docs at once. It surfaced 19 tickets filed in two months, flagged that the feature had been partially built and abandoned, and gave verbatim quotes with sources. The value is not the summary. It is not duplicating work someone already started.
8. Automate execution, keep decisions human - Defining the feature, the API surface area, and the abstraction layer stay hands on. Code writing is the part that runs on its own. The PRD skill is explicitly instructed to challenge the PM's assumptions rather than accept them.
9. Agents are already the majority user, so validate for them - Agent evals spins up a sandbox, gives an agent a real task against the product, and watches. It caught that agents could not find the fine-tunable models page because it was not linked from the quick start. Dozens of docs fixes came out of this. Charles calls agent success the new bar for UX.
10. They refused to oversell the gains - No story points, so no proof, but velocity is up more than 5%. Charles finds 3x claims suspicious, because discovery, debate, and coordination do not get magically better with AI. Costs stayed sane partly through open weight models, partly because optimizing for collective output never produced the runaway token budgets others report.
Go Deeper
As promised, I put together this section to help you become the best version in your product management journey. If you’re starting from zero on this, begin with How to build a Team OS in Claude Code with Hannah Stulberg, then scale it up with How to build a Company Operating System with Hermes and OpenClaw. For the skills layer specifically, 3x CPO Oji Udezue on the Essential Claude Skills for PMs is the one to read. And if section 3 sold you on agent evals, go to How to Build Frontier-Lab Quality Evals with Daniel McKinnon. Finally, Pavneet’s take on the one-page PRD lines up almost exactly with what Srini Raghavan showed from Freshworks.
You can get my podcasts on Apple, Spotify, and YouTube.
If you want to advertise, please email - productgrowthppp[at]gmail[dot]com.