Saw this earlier but just getting back to it now... basically I went through three laptops during my undergrad because I kept underestimating my workload. If you are gonna do Docker or run a few VMs simultaneously, 16GB is the bare minimum and it feels cramped fast. I remember trying to run a local microservices environment on 16GB and my system literally ground to a halt. Regarding the GPU, unless you are planning on doing really heavy local model training, you probably dont need a beefy dedicated chip. Most ML projects in school are small or you will just offload them anyway. To answer your battery question, it really depends on the architecture of the chip. Heres how a few options usually stack up for us:

• Apple MacBook Pro 14 M3 Pro 36GB RAM 512GB SSD Pros: The battery life is actually insane, you can legit go a full day without a charger. Unix is native. Cons: High entry price and no real hardware upgrades later.
• Lenovo ThinkPad P14s Gen 5 AMD Ryzen 7 32GB RAM Pros: Best keyboard in the game and runs Linux like a dream. Much more cost-effective for the specs. Cons: Battery life is just okay, maybe 5-7 hours under load.
• Dell XPS 15 9530 Intel i7-13700H 32GB RAM RTX 4050 Pros: Great if you need CUDA for specific ML projects. Cons: It gets pretty hot and the battery drains fast when the GPU kicks in. Honestly, prioritize 32GB RAM over a fancy GPU. Your IDE and containers will thank you way more over the next four years.

Stumbled on this thread and wanted to jump in with a more budget-conscious perspective. Tbh, a lot of CS students overspend on their first machine thinking they need a powerhouse, but you can get way more value if you look at the right Windows or Linux-friendly hardware. Regarding RAM, 16GB is definitely the baseline now. If you are planning on running multiple Docker containers or heavy VMs, 32GB is nice to have, but here is a tip: look for a laptop with SODIMM slots so you can upgrade it yourself later for cheap. Buying a machine with 32GB pre-installed often comes with a massive manufacturer markup. For the GPU, unless you are doing heavy local model training, you probably dont need a dedicated card. Most ML coursework uses cloud resources like Google Colab or school clusters anyway. You are usually better off saving that money or putting it toward a better screen or more storage. If I were starting today on a budget, I would grab something like the Lenovo ThinkPad E14 Gen 5 AMD Ryzen 7 16GB RAM or even look for a refurbished HP EliteBook 845 G9 AMD Ryzen 7 32GB RAM. These are absolute workhorses, have great Linux support, and are built to survive being shoved in a backpack every day. Plus, the keyboards are way better for long coding sessions than most thin-and-light laptops. Save your cash for extra monitors or coffee... trust me, you will need the coffee more than a high-end GPU.

Quick reply while I have a sec... honestly I have gone through so many laptops over the years and the one thing that always kills the vibe is when a small part breaks and the whole machine becomes a paperweight. For a four-year degree you want something that isnt just fast but actually stays in one piece. I switched to machines that are a bit more rugged or at least user-serviceable because having your only dev environment die during finals week is a nightmare I wouldnt wish on anyone. Personally I found that reliability matters way more than having the flashiest chassis.

• If you want something that will legit last you till graduation and beyond, look at the Framework Laptop 13 DIY Edition Intel Core Ultra 7 155H 32GB RAM 1TB SSD. The fact you can swap out ports or fix the screen yourself is a lifesaver when you are a broke student.
• For pure tank-like reliability, the Lenovo ThinkPad P14s Gen 5 AMD Ryzen 7 PRO 7840U 32GB RAM 1TB SSD is basically the gold standard for CS students who want to run Linux. Those keyboards are a dream for long nights of debugging. Definitely push for 32GB tho... Docker is a resource hog and you dont want to be fighting your OS while trying to code. Integrated graphics are totally fine for 90 percent of what you will do in school, so dont stress the dedicated GPU unless you are hardcore into running local LLMs or heavy rendering.

I totally agree with the advice you mentioned about Unix-based systems being the gold standard for software development. Having that native terminal experience is a massive time-saver when you are setting up dev environments or managing remote servers. Honestly, I spent way too many years troubleshooting OS-specific path issues before making the switch, and I really wish I had done it sooner. To add one small point that often gets overlooked: really look at the screen and keyboard quality. You are going to be staring at code for hours on end, so a high-resolution display with good brightness is worth every penny to avoid eye strain. On your specific questions:

• Definitely aim for 32GB of RAM if the budget allows. Between running multiple Docker containers, a heavy IDE, and fifty browser tabs, 16GB fills up way faster than you would think.
• Dont worry too much about a dedicated GPU for machine learning. Most of your heavy training will likely happen on university clusters or in the cloud anyway.
• Battery life is huge. You dont want to be that person hunting for a wall outlet in the middle of a three-hour lab session. I have seen so many students buy a bulky gaming laptop because of the raw specs, only to regret it by sophomore year because it dies in ninety minutes and weighs a ton. Focus on that sweet spot of portability and sustained performance. If you get a machine that handles your multitasking well now, it will easily see you through to graduation without feeling like a relic.

> Is 16GB of RAM really enough, or should I push for 32GB if I plan on running multiple Docker containers or virtual machines? This situation definitely reminds me of a specific case involving a colleague who was choosing between the Dell and Apple ecosystems. He approached the decision with extreme precision, basically spending a month conducting a methodical comparison of chassis rigidity and key travel distance to ensure he selected the optimal development environment. After he finally settled on a high-end unit, he ran into an unexpected issue where the specific chemical finish on the palm rest caused a persistent allergic reaction on his wrists. He spent the next two weeks researching specialized vinyl skins and ergonomics kits just so he could operate the machine for more than ten minutes at a time. It turned into a massive ordeal with the manufacturers technical support department since they had no documented cases of that specific material interaction. Its pretty wild how you can plan for every technical specification like RAM or storage, but then something completely random like that happens and renders the entire hardware debate irrelevant.