Table of Contents
Alright, here’s your weekly tech roundup—no fluff, just the stuff I’d actually want to know. This week’s big theme? Apple pushing deeper into AI on-device, plus some serious hardware competition in the AI chip world. And yes, there’s also a reality check on what chatbots can (and can’t) do.
- Apple’s “Glowtime” event teases iPhone 16 and a bigger push into AI features.
- Cerebras shows off an AI chip with over a million cores—positioning it as a serious alternative to Nvidia.
- Mike Krieger from Anthropic talks about Claude’s strengths, limitations, and why “useful” matters more than “smart.”
If you were hoping Apple would keep AI “mostly in the cloud,” this event wasn’t that. What I noticed in the coverage is how much emphasis there is on putting AI features closer to the device—faster responses, more privacy, and fewer “wait, I’m sending this somewhere” moments.
- Apple
- At Apple’s “Glowtime” event, the headline is the iPhone 16—and the company’s next wave of AI features. The vibe is pretty clear: Apple wants AI to feel less like a separate app and more like something built into everyday tasks.
- In practical terms, I’m looking for the stuff that saves time without getting in the way. Think better photo and video handling, smarter suggestions in places you already use on your phone, and AI that can actually understand context (not just keyword-match). The reports also point to updated iPhone models, potential new color options, and the usual “small hardware tweaks” that end up mattering day-to-day—like performance, battery efficiency, and camera behavior.
- Will it be magic? Probably not. But if Apple nails the “quietly helpful” part of AI—like rewriting a message, organizing content, or summarizing something while you’re busy—that’s where it gets real.
- Cerebras
- This one’s for the hardware nerds (and honestly, for anyone who’s tired of Nvidia being the default choice). Cerebras is showing an AI inference chip with over a million cores, plus 44GB of high-speed memory. The claim is that it can deliver “instant” AI processing—basically aiming to cut the latency that makes AI feel clunky.
- Why does that matter? Because in real life, speed isn’t just a benchmark number. If an AI system responds in 0.2 seconds instead of 2 seconds, you stop waiting and start using it like a tool. You can run it more interactively—customer support, search, real-time personalization, and other workloads where delays kill the experience.
- Now, I’ll be honest: “million-core” sounds wild, but what I’d want to see next is how it performs on real deployments. How stable is it under sustained load? How does it compare on cost per inference? And what’s the developer experience like? Still, it’s a legit challenge to Nvidia’s position.
- Anthropic
- Mike Krieger (Anthropic’s new Chief Product Officer) had a pretty grounded take on AI chatbots like Claude. What stuck with me is the emphasis on usefulness—not just impressive text generation.
- In my experience, this is where a lot of chatbot demos fall apart. Sure, it can sound confident. But can it handle messy prompts? Can it stick to constraints? Can it avoid making stuff up when you ask for specifics? Krieger’s point lines up with what many teams learn the hard way: you have to design for real problems, real workflows, and real user expectations.
- He also connects it to his earlier work at Instagram, which makes sense. Social apps live or die by how well they support everyday tasks. If Anthropic is applying that mindset to AI products, that’s a good sign.
I’m picky with “new AI tools,” mostly because a lot of them sound great and then disappoint in the details. These are at least interesting, and each one targets a pretty specific use case.
- BypassGPT.co— Transform text made by AI into writing that feels more human, avoiding detection by tools such as GPTZero
- If you’re using this for legitimate rewriting (like improving tone or clarity), it can be useful. Just don’t treat “bypass” as a magic wand—quality still matters, and you’ll still want to review what it outputs.
- Video Studio AI— Create great videos from words or pictures that show real facial expressions
- What I’d test first is how it handles consistency—same character, same face, same vibe across multiple clips. Facial expression quality is one thing; keeping it coherent is another.
- Vocal Remover— Remove the singing and let the music play using good quality backing tracks
- For vocal removal, the real question is how clean the separation is when the vocals are mixed strongly with the instruments. If you’ve ever tried one of these tools and ended up with “robotic artifacts,” you know what I mean. I’d run a couple tracks you know well before trusting it for anything important.
Here’s today’s prompt. If you want to make it easier, pick a niche you actually care about (or one you’ve worked in). You’ll get a better output.
Create a comprehensive marketing strategy for a [insert niche] business. Include the following elements: target audience analysis, unique value proposition, key marketing channels, content marketing ideas, social media strategy, and a budget breakdown. Additionally, suggest potential metrics to measure the effectiveness of the strategy.
Quick tip: when you do the budget breakdown, try assigning a percentage (like 30% content, 25% paid ads, 15% email, etc.). It forces the strategy to feel real instead of just “nice ideas.”
