EDITION 07:00 IST
ARCHIVERSS WIRE
Press and hold. Everyone else tells you what happened.
The Daily Five

Samsung sets Unpacked for July 22, Gemini 3.5 Pro lands, and everyone starts making their own AI chip

Foldable season is officially here — Samsung confirmed a July 22 Unpacked with three foldables. Meanwhile Google's Gemini 3.5 Pro is reportedly shipping, and the AI giants are quietly turning into chipmakers. Plus a hidden iPhone trick and a camera myth worth busting.

  1. commentary

    Samsung Unpacked is official: July 22

    It's official — Samsung's next Unpacked is July 22, and the lineup is stacked: Galaxy Z Fold 8, a new Z Fold 8 Ultra, the Z Flip 8, plus Watch 9 and Watch Ultra 2, with smart glasses rumored to make an appearance. Three foldables at one event is Samsung going all-in. But the real story isn't the hinge or the vanishing crease — it's how deep Galaxy AI is baked into everything. Foldables have quietly stopped being a gimmick. 📱

    The problem
    Samsung has historically shipped one foldable line update per cycle — a single Fold and a single Flip — leaving a gap for anyone who wanted more differentiation between models.
    What changed
    Samsung's July 22 Unpacked lineup is now confirmed at three foldables in one event — the Galaxy Z Fold 8, a new Z Fold 8 Ultra, and the Z Flip 8 — plus Watch 9, Watch Ultra 2, and possibly smart glasses, with Galaxy AI baked deeply into all of them.
    The catch
    More devices in one event means more ways to overpay for the wrong one — with an Ultra tier now in play, picking the right foldable for your actual needs matters more than it used to.
    What it means for you
    Don't default to the top-tier Ultra out of habit — the real story of this Unpacked is how deep Galaxy AI runs through the whole lineup, so figure out which AI features you'd actually use before picking a price tier.
  2. ai

    Gemini 3.5 Pro is reportedly here

    Google's Gemini 3.5 Pro is reportedly rolling out (targeted for July 17) with a 2 million token context window, a Deep Think reasoning layer, and autonomous workflow abilities. Note the word reportedly — the date is from leaks and reports, not an official Google post yet. If the specs hold, a 2M context window is the headline: you could feed it entire codebases or books in one go. Worth watching, not worth pre-ordering your hype. 🤖

    The problem
    AI labs routinely leak or tease upcoming models well before anything is officially confirmed, making it hard to know what's real news versus hype building ahead of a launch.
    What changed
    Gemini 3.5 Pro is reportedly rolling out around July 17 with a 2-million-token context window, a 'Deep Think' reasoning layer, and autonomous workflow abilities — specs that, if accurate, would make it the largest-context production model available.
    The catch
    The word 'reportedly' matters here — this is coming from leaks and reports, not an official Google announcement, so the date and specs could still shift.
    What it means for you
    If the 2M context window holds up, it's worth revisiting for any workflow involving entire codebases or long documents — but hold off on planning around it until Google actually confirms the release.
  4. tip

    Your iPhone has Shazam built in

    You don't need a separate app to identify a song. iPhone has Shazam built into Control Center. Go to Settings, Control Center, add Music Recognition, then tap it any time a song is playing and your phone will name it — even with headphones in. It quietly logs everything it finds, so you can scroll back later. One of those features Apple never bothered to advertise. 💡

    The problem
    Identifying a song playing around you usually means fumbling for a separate app, and Apple never clearly advertised that it built this exact feature directly into iOS.
    What changed
    iPhone has Shazam built into Control Center — add Music Recognition under Settings, Control Center, and tapping it identifies any song playing nearby, even through headphones, while quietly logging everything it finds for later.
    The catch
    It only works if you've added it to Control Center first — it's not visible or active by default, so most people never discover it exists.
    What it means for you
    Add Music Recognition to your Control Center today — next time a song catches your ear in a cafe or a car, you'll have it identified and logged in one tap.
  5. myth buster

    More megapixels won't save your photos

    Every launch season the megapixel numbers climb, and every year they matter less than the ad makes you think. Most 200MP phones bin those pixels down to around 12MP for your actual shots. What genuinely moves the needle is sensor size, individual pixel size, and the computational photography doing the heavy lifting after you hit the shutter. A well-tuned 50MP camera beats a lazy 200MP one every single time. Read the sensor, not the sticker. 📸

    The problem
    Camera marketing leans hard on megapixel counts every launch season, training buyers to equate a bigger number with a better photo — even when that's rarely how image quality actually works.
    What changed
    Most 200MP phones bin those pixels down to around 12MP for your actual shots — what really moves the needle is sensor size, individual pixel size, and the computational photography processing the image after you hit the shutter.
    The catch
    A high megapixel count isn't useless — it helps with cropping and certain zoom modes — but it's not the number that determines whether your everyday photos look good.
    What it means for you
    Next time you're comparing phone cameras, look past the megapixel figure on the spec sheet and toward sensor size and processing — a well-tuned 50MP camera will beat a lazy 200MP one every time.
  6. hot take

    Everyone's building their own AI chip now

    Anthropic is reportedly in talks with Samsung to build a custom chip tuned for its Claude models, while OpenAI and Broadcom's Jalapeño inference chip hit tape-out in just nine months. The pattern is impossible to miss: Google, Amazon, Meta, OpenAI and now Anthropic all want their own silicon. The takeaway — the real AI moat isn't just the model anymore, it's owning the hardware it runs on. Whoever controls performance-per-watt controls the economics. ⚡️

    The problem
    Every major AI lab depends on renting compute from someone else's chips, which means their margins and their roadmap are ultimately controlled by whoever makes the silicon they run on.
    What changed
    Anthropic is reportedly in talks with Samsung to build a custom chip tuned for Claude, while OpenAI and Broadcom's Jalapeño inference chip hit tape-out in just nine months — following Google, Amazon, and Meta, who already build their own silicon.
    The catch
    Designing and shipping custom silicon takes years and enormous capital — announcing talks or hitting tape-out is a long way from having chips that actually outperform renting from Nvidia at scale.
    What it means for you
    If you're picking an AI vendor for the long term, pay attention to who's investing in their own chips — performance-per-watt, and therefore price, increasingly depends on who controls the hardware, not just who has the best model today.