Dezine Zync

During my Bengaluru trip last month, I came across a significant disadvantaged with my then Macbook setup. It’s a 14” M1-Pro top of the line with 32GB of RAM.

For my day-to-day work, it did a fine job, no complaints. Until I realised, that was only the case when it was plugged into the Studio Display at my home desk. The moment it was off the power cable (thunderbolt in my case), it would immediately show its age and current capacity to do things.

Frustrated by this, and not willing to wait any longer for Apple to ship a M5 Max capable Mac Studio, I decided to pull the plug and get myself a M5 Max MacBook. Now I wasn’t going to make the same mistake as a few people did ordering a M5 Max in a 14” chassis, and simultaneously, I wasn’t sure I wanted a 16” MacBook as my daily driver. But seeing as I had very few options, I ordered a 16” M5 Max 18-core CPU, 40-core GPU, with 48GB RAM, and closed my trusty old M1-Pro for the day.

I forgot to check the delivery date on this, and since it wasn’t blocker for my present work, I didn’t think much of it, and carried on. Two days later, it had already arrived! Luckily for me, this was a Sunday, so I immediately got it hooked up to my M1-Pro MacBook using Thunderbolt, and began migrating my data.

The Mac

Macs have been my favourite and preferred computing devices for almost two decades now. What began as an aspiration is now a necessity for me, simply because of:

• My workflows which have evolved over the many years to use macOS and the software available on the platform.
• My muscle memory with all the keyboard shortcuts, text rendering and legibility
• And of most importance: I get to use a 5K Studio Display 😉

And these machines are a delight to hold in the hand. While some people take extra caution when carrying them around, I would carry, hang, lightly drop my M1-Pro MacBook on a table without a second thought. I know how well these parts are built, and that extra caution some people take doesn’t bother me: I don’t do it because it’s extra mental work for me.

They are, also, a delight to look at! The smooth corners are obviously the people’s choice award winners, but my surprise was the Silver colour which is new to me! My brother has a M3-Pro MacBook 14” in Space Black, and I’m not a fan. I’ve experienced the MacBook as a silver laptop for far too long, and I’ve just gotten used to that. It also pairs nicely with the Studio Display. This silver and the M1-Pro’s silver are very different, and not slightly, but obviously.

What wasn’t a surprise, is how heavy the 16” MacBook is with a larger heatsink for the M5-Max.

The Work

As I was able to transfer all my existing accounts, data, and software within an hour using a Thunderbolt 4 cable, I was ready to get back to work on the following Monday. And everything just felt normal until I booted up my usual software to work in, in this case, Xcode.

I instantly saw 3-5x improvements in compile times for clean builds, and up to 8x improvements for incremental builds for both Pockity’s iOS and macOS targets. Furthermore, I saw a solid 25x improvement when training Pockity’s NLP CoreML model. So this meant I needed to try training a larger dataset now that I have a machine which won’t melt its internals attempting it. And behold, the obvious: not only did it not melt, it maintained the consistent performance bump I saw with the older dataset from March, 2026.

The end result was a slightly larger model, now fine-tuned to handle a larger variety of how people type in their expenses.

Here’s a side-by-side comparison of the model that’s shipping in v2026.5 of the app, and the newer one which will ship in v2026.7:

Now that I have a more capable machine, I’m already excited to explore more CoreML and CreateML projects to bring more invisible machine learning benefits to my customers.

Something else that caught me off guard is how fast Apple Intelligence is on these newer chips. Apple Intelligence launched nearly 2 generations after the M1 series chips were introduced, so they clearly weren’t optimised for running Apple’s models. The M5 series chips, especially the M5 Max is wicked fast! In April 2026’s release of Pockity, I announced MCP and AppleScript support for the Mac app, and wrote about using Shortcuts Automation + Apple Intelligence to automatically ingest expenses into the app from SMSes and emails. I was using Apple’s private cloud compute option in that, the screenshots still reflect this, but when I tried it with the M5 Max chip, it was near instantaneous. If you’ve built such shortcuts before, the app highlights each step when it’s working on it. The M5 Max immediately highlights the next row, and not by error; because it has completed that step so quickly, the animation never gets a chance to play out completely 😄

---

And finally, there’s also my colour grading and Youtube work for which I use Resolve Studio and FCP 11. I was expecting a huge bump in encoding performance here as well, but it wasn’t immediately evident to me. Maybe because these are MP4 containered H265 files from the DJI Mini 4 Pro, so I’ve yet to try it with ProRes files and even true 12-bit files. So I’ll be sure to do a follow up post of this later.

While this video is not demanding, as you can clearly see, I also tested some of my complex colour grading projects in Resolve, which unfortunately I cannot share here due to *legal agreements**. I can say for certain, the lag free colour grading experience is truly a sight to behold, especially with 6K files, and lots of parallel nodes trying to hammer the GPU as it casually hums along.

The Lifestyle

This is a very heavy machine. I’ve never owned a laptop this big or heavy in the past, and it took me two weeks to get used to the larger surface for resting my hands and palms. Since the keyboard and trackpad mostly stay the same between the 14” and 16”, that was a non-issue. But getting used to the larger display also took a minute. I had to go around in all the software I use, especially Xcode, and tweak the font sizes so they are comfortable and legible for me at this new distance that I now need to hold this giant display at.

Then there is the new silver colour (new to me)! It is a 100% improvement over the gray of the M1 series MacBook pros. It once again feels like a MacBook Pro, and I don’t know what the M1 series were trying to be. I wonder why Apple did not call it “Starlight” though. It sounds cool compared to just “Silver” 😄

I’ve yet to figure out a good backpack or handbag for the new laptop since my existing one clearly won’t fit it. I’m open to recommendations, but it should be available for purchase directly in India. I am still unsure if I’ll carry this when I travel, or continue to carry the M1-Pro when I do since now I no longer need it do so much of the heavy-lifting.

The LLM Trap

If you were wondering if I am going to write about LLM models and running them locally on my Mac: The simple answer is, “no”.

Here’s the long version if you’re interested: I did the math of running LLMs locally. I don’t even use them much. But as an experiment, I tried in March and April, to use two different ones: March was Gemini, and April was Claude. In both months, I definitely reached out to use them, but I cannot get myself to use them as I see a lot of people do: organise and plan a task, then let the models do all the typing. Regardless, I then tried a local model. Which one is not relevant, however my findings were! When running these models locally, what I have not seen anyone else account for are two things:

• power consumption for the computer hardware
• power consumption for cooling the room that hardware is in

Now I live in a place which is hot for six months of the year, pleasant for the remaining six. Cooling is essential. And after considering those two, it’s far far cheaper for me to use a hosted model should I need it. Their plans are inexpensive too when I compare it to cooling the room plus the extra energy costs consumed by the devices. It’s a net-loss setup for me.

I don’t suspect most people have done that math, but I’m glad I did. It made my decision to buy the 48GB model very easy, as compared to the higher RAM specifications, because it simply isn’t worth it for me.

Looking forward…

What I’m curious to try next and explore is Davinci Resolve’s upcoming photo editor in v21! I have a feeling it’s really going to unlock new potentials for this Mac and me, and what we can do together.

As I mentioned earlier, I’m also looking forward to exploring new and more CreateML and CoreML mini-projects to integrate into my apps. I’m already working on new Mac software where having this capability will be very useful.

So as I continue to wrap my head around this incredible hardware, I’ll skip the closing notes, and leave you with this beautiful photo of it: