I watched three different WWDC recaps back to back this June, and by the third one I noticed something: nobody was summarizing the keynote platform by platform anymore. That's usually how these recaps go — iOS, then watchOS, then macOS, in order, like reading a table of contents out loud. This year, people were describing it in themes instead. That's not a small shift. It means Apple actually restructured the talk, not just the slides.

That's the thing about "best presentations of the year" lists — they're rarely about who had the shiniest deck. They're about who made a structural decision that changed how the audience processed the information. 2026 gave us a couple of genuinely useful case studies on that front, so let's look at what actually happened on stage, not just what got announced.

Apple's WWDC 2026: breaking its own format on purpose

Apple's WWDC keynotes have followed the same skeleton for years: walk through each operating system one at a time, closing with a developer-facing segment. At WWDC 2026 — Tim Cook's final keynote as CEO before handing off to John Ternus — Apple dropped that structure entirely. Instead of iOS, then iPadOS, then macOS, the keynote was organized around three themes: platform improvements, trust and safety, and Apple Intelligence with the reworked Siri.

The stated logic, and the one that held up watching it, was that the same features were landing across iPhone, iPad, Mac, and Watch simultaneously, so presenting them per-device had started to feel arbitrary. When one feature touches five products, walking through the products one at a time means repeating yourself five times or awkwardly cross-referencing a segment the audience hasn't seen yet. Organizing by theme instead of by device solved a real information-architecture problem, not just a pacing one.

I'd also point out the more than ten minutes spent on child safety and parental controls — mandatory child accounts under 13, granular app permissions, an "ask to browse" gate in Safari. That's an unusual amount of runway for a topic that isn't a flashy feature reveal. In my experience, when a company gives a "boring" but important topic that much stage time, it's rarely accidental — it's a deliberate signal that the topic matters enough to slow the whole keynote down for it, regulatory pressure or not.

One more visible change: the camera work moved away from the heavily stabilized steadicam look toward more visibly handheld footage, giving the keynote a rougher, more human feel. Small production choice, but it changes the emotional register of the whole talk before a single slide even shows up.

NVIDIA's GTC 2026: the founder-narrator format at scale

Jensen Huang's GTC 2026 keynote in San Jose ran close to two hours, and it's worth studying for the opposite reason Apple's is: it barely resembles a traditional slide deck at all. It's closer to a factory tour narrated by the person who built the factory. Huang walked the stage making one sustained argument — that AI has moved from an experimental cost center to an industrial production system, with tokens as the literal output.

Structurally, that's a big bet. Most keynotes chain together a series of announcements. Huang's kept returning to the same throughline — compute as a convertible, revenue-generating input — and used each announcement (the Vera Rubin platform, the NemoClaw agent framework, new Nemotron and Cosmos models) as supporting evidence for that one argument rather than as a standalone headline. It's the same trick a good persuasive essay uses: state the thesis early, then spend the rest of the time proving it from different angles.

The trade-off is real, though. A two-hour, single-narrator format asks a lot of an audience's attention, and it only works because Huang has enough stage command and enough genuinely new material to sustain it. I wouldn't recommend this structure for most corporate presentations — it's a format that scales with how much the speaker themselves is the draw, and most of us aren't presenting to a sold-out arena.

What's actually transferable to a normal deck

Neither of these was really about slide design in the traditional sense — better fonts, cleaner charts, tighter grids. The lessons sit one level up, at the structural layer, which is honestly where most decks quietly go wrong before anyone even opens PowerPoint.

  • Organize by argument, not by inventory. Apple's shift from "here's every product update" to "here's what we're trying to say" is the single most reusable idea here. If you're presenting quarterly results, for example, resist the pull to walk department by department. Group around the two or three things you actually want the room to remember.
  • Give your "unglamorous but important" section real time. The child-safety segment worked because Apple didn't rush it. If there's a compliance update, a risk disclosure, or a process change buried in your deck, don't apologize for it with 30 seconds and a bullet list — give it room, or cut it entirely.
  • One throughline beats five headlines. Huang's keynote holds together because every new product answers the same question. Before you build slide two, it's worth writing down, in one sentence, the single idea every other slide needs to support.
  • Match format to how much attention you've actually earned. A two-hour narrator-led keynote works for Huang because of who he is and what NVIDIA has to show. For most internal presentations, 20 tight minutes beats 60 rambling ones — know which category you're in before you plan the runtime.

Where I'd push back

Not everything from these keynotes is worth copying, and I think it's a mistake when "best presentations of the year" pieces treat big-company keynotes as universally aspirational. A theme-based restructure only makes sense when your content genuinely crosses categories — if your quarterly update really is five unrelated department reports, forcing them into three artificial "themes" will read as gimmicky rather than clarifying. And the founder-narrator format leans hard on the founder actually being compelling live; borrowing the structure without the stage presence usually just produces a long, underprepared talk.

The common thread, if there is one, isn't a visual style. It's that both of these keynotes made a structural decision before they made a design decision — they figured out how the information should be organized, then built the visuals to support that, instead of the other way around.

If you're rebuilding a deck this year, that's probably the more useful question to sit with than "what fonts is everyone using now": what's the one thing you want the room walking out remembering, and does your current structure actually point at it?