hi bud buds,
last week, microsoft announced it will rebuild windows from the ground up to be “ai-native”, with voice-to-text as a major bet. the product vision: talk to your computer, it understands you, then it gets things done for you.
“in our minds, voice will now become the third input mechanism to use with your pc. it doesn’t replace the keyboard and mouse necessarily, but it’s an added thing and it will be pretty profound and a new way to do it.”
i personally love this product direction. i already use voice-to-text constantly (shoutout willow), and recommend every product manager to do the same. it helps me get my initial thoughts out 10x faster before polishing them. in a survey at my company, voice-to-text ranked just behind first-draft doc generation as the most-used ai tool among product managers.
however, this announcement also intensifies an already crowded platform war, where ai copilots are being built at every layer of the consumer tech stack:
in operating systems, with windows (and soon macos) going ai-native
in applications, with app developers launching in-app copilots
in chatbots, with openai enabling apps inside chatgpt just last week
in the near future, users may have 2-3 copilots stacked on each other: inside their os, browser, and web app simultaneously.
so who wins when every layer wants to be your ai assistant? imo, the operating system. device makers can push their copilots more aggressively, and os-level assistants can leverage context across apps, two advantages that no browser or app can match. to illustrate this point:
imagine a superintelligent version of notion that can do anything inside notion perfectly
now imagine a superintelligent siri that can navigate across all your productivity apps, pulling whatever data it needs, and taking actions anywhere on your device it needs to
the latter feels way more powerful.
that said, if consumers end up using a mix of ai-enabled devices (smartphones, earbuds, glasses) from different companies, model developers like openai and anthropic could win by offering continuity across ecosystems. a good personal comp for this is lastpass (my password manager), which i use across multiple devices and browsers. i like that it just works seamlessly everywhere.
notably, i don’t think app-level copilots are well-positioned in this war unless their workflows are highly isolated. for example, for most productivity workflows which span multiple apps (e.g. spreadsheets, docs, presentations), os-level assistants have a clear natural advantage: they’re the only assistant that can truly sit above every app.
thus, if you’re a pm building an app-level copilot right now, think carefully: in a world where your copilot sits beneath two or three others, what’s your right to win?
have a great week, y’all. i love you.
- dj
last week’s biggest product releases
big news! huge!
salesforce launched agentforce 360, a platform for creating and managing agents with access to sales records, tableau visualizations, and salesforce governance / permissions. so many sales + rev ops activities seem ripe for ai automation, and salesforce is the source of truth for sales context, so i’m bullish on any bets salesforce makes on agents
walmart announced a partnership with chatgpt, enabling users to browse and purchase walmart products within chatgpt. when the world’s biggest retailer teams up with the world’s most disruptive tech company… watch out.
facebook re-released job listings (after deprecating them three years ago) for entry-level, trade, and service roles. with openai also recently announcing its own jobs platform, linkedin is seeing growing competitive pressure from multiple tech giants at once
personal faves
amazon released virtual healthcare for kids, starting at $29 per visit with no insurance required, covering common conditions and rx renewals. i love that this could make healthcare more accessible for families without strong insurance or the flexibility to take time off work
uber released digital tasks, letting drivers earn extra income by completing quick data-labeling jobs in the uber driver app. i love this: just a clever new solution to serve a nearly identical icp, problem, and jtbd
doordash announced waymo robotaxi deliveries in phoenix. customers will retrieve orders from the trunk via the doordash app
google released help me schedule for gmail, an ai tool that suggests meeting times and inserts suggested time slots into emails
nice job guys
apple released m5 chips in the new ipad pro, macbook pro, and vision pro, boosting cpu, graphics, and ai performance
anthropic released skills for claude, letting users build customizable task folders to help claude better handle specific work tasks
google released new ai editing tools for flow (its ai filmmaking tool), including the ability to add / remove objects, restyle scenes, and extend video length
coinbase released coinbase business, a platform enabling companies to send and receive stablecoin payments
pinterest released genai controls, letting users limit ai-generated content in their feeds
spotify released text instructions for its ai dj, after previously only supporting voice commands
youtube released a new video player, with a cleaner and more immersive design (seemingly inspired by liquid glass)
google released recovery contacts, allowing users to add trusted contacts for account recovery
threads released group chats with up to 50 followers
kayak released ai mode, an ai chatbot for trip planning and booking
other llm news
anthropic released claude haiku 4.5, a small model with near-frontier coding performance at reduced cost and increased speed
google released veo 3.1, an ai video model with improved audio, prompt accuracy, and scene textures
microsoft announced mai-image-1, its first in-house text-to-image model
idk tbh
openai announced erotica for adult users, relaxing safety restrictions for "verified adults" to engage in erotic conversations. this feels like a huge liability, incredibly brand dilutive, and overall an unwise product direction
google released hidable sponsored results, now grouping sponsored results together in a collapsible section. this seems like it would be significantly revenue dilutive, but i’m sure they’ve done extensive testing? i’m just not sure what they gain from this…
facebook released photo edit suggestions, allowing ai to proactively suggest photo edits from your camera roll. feels both creepy and not very useful… a fantastic combo
google released ai-powered makeup feature, allowing google meet users to apply virtual makeup during calls. this feels weird to me… but i’m definitely not the target user
that’s it for this week; thanks for reading. if you enjoyed this issue, please consider sharing it with a friend so i can get rich and fly to visit my childhood friends more often in san diego.
love, dj