hi bud buds,
last week, i wrote about gpt-5 feeling like a disappointing step forward, and a damning blow for all that “agi” yap:
“i’ve always been skeptical of sama’s claims that “agi” is imminent, and this feels like further validation that we’re still a long way off”
since then, similar critiques have hit the mainstream:
new yorker: what if ai doesn’t get much better than this?
financial times: is ai hitting a wall?
what does this news means for product managers? my current read:
gpt-5 flop signals llm progress is rapidly decelerating
this is extremely bad for frontier labs (open ai, anthropic), who are burning billions to marginally improve models
meanwhile, cheaper / oss substitutes will offer similar performance at a fraction of the cost
because of this, the new winners of this next ai era will be incumbent software businesses
🌶 therefore, leaving your pm role to join that hot ai startup in your dms might not be the play
why incumbents specifically?
if ai isn’t getting 10x better, humans will need stay in the loop, and the co-pilot model will beat out the agentic model. therefore, ai startups touting “ai-native” or agentic design will be a net-negative for users. the ideal architecture will be bolting ai features onto existing platforms, augmenting workflows instead of replacing them
with inference and dev costs heading to zero, nothing stops incumbents from rapidly cloning your startup idea as a feature and giving it away for free to their entire existing customer base
the caveat: who the hell knows? shoot your shot! life is short, and startups are fun! but don’t say i didn’t warn you…. 🌶 👀
have a great week, y’all. i love you.
- dj
last week’s biggest product releases
big news! huge!
amazon launched same-day grocery delivery in 1,000+ us cities; free for prime orders over $25. this ramps up competition with instacart and walmart
personal faves
airbnb launched reserve now, pay later, letting users book without upfront payment and making cancellations easier if plans change
google launched preferred sources, letting users elevate specific sources in search results. excited to enable both wsj + nyt for more thoughtful (?) takes across the aisle, though net-net this probably worsens ideological echo chambers 🥴
nice job guys
anthropic launched conversation ending capabilities to claude, letting models terminate harmful or abusive chats
apple launched ios 26 beta 6 with new ringtones, faster app launches, and small liquid glass tweaks. i was a liquid glass hater at first (felt distracting), but after beta testing, i’m neutral. imo it’s fine / almost entirely immaterial
google calendar launched event duplication using ctrl-click, a public feature request from stripe’s john collison
waymo added spotify integration, letting robotaxi riders stream music directly
other llm news
nvidia launched cosmos reason, a new 7b “world model” for robots + physical ai use cases
anthropic extended claude’s context window to 1m tokens (~750k words / 75k lines of code)
idk tbh
apple demoed allowing siri to control apps with voice commands only. siri is an absolute travesty, so i’m pro anything to improve her
google flights announced natural language search. imho, i expect low adoption for this. flight search criteria are very concrete (dates, prices), which existing filters already handle efficiently. imo, natural language shines more in abstract searches, like asking netflix for a ‘vibey sadboi movie to watch while i’m hungover and have the sunday scaries’
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 buy a $300 ralph lauren jacket i’ve been eyeing.
love, dj