WSJ refers to 2026 college grads as "AI native". I don't really know about that; to me, native would be truly imply having grown up with it moreso.

In any case, this crop entering the workforce (theoretically) would have had ample time during between/for class to use and ramp on AI, especially compared with those already used to or busy doing things the older fashioned way at $dayjob, however much AI crept or jolted in. So, I guess the moniker fits, comparatively.

Native in terms of AI being already in full-ish force when entering the workplace, perhaps.

Curious how HN thinks this matters, if anything.

Comments URL: https://news.ycombinator.com/item?id=48276079

Points: 4

# Comments: 1

2. Go to a Meetup or Luma events for your industry of interest near your city and talk to real people about what their company does and how your experience making projects relates.

3. Catch up with people you know in real life from your studies about the places they or their contacts work at and your interest in an internship.

It probably starts with some kind of connection.

Then when you talk to a person at the company who needs/wants an intern (the "hiring manager"), you need them to like you. Be nice to them. Listen attentively. Be professional, and also be yourself. Focus on how you can be useful for their team/project/company, not on what you need.

After such communications, follow up with a "thank you" email (or text if familiar) with specifically what you enjoyed about talking to them.

Make sure your resume is perfectly formatted, grammatically correct and not AI generated. Edit it yourself to retain your own voice. You can include skills, technologies, passion projects, and education on your resume, even if you lack work experience.

  uv pip sync requirements.txt

  python3 -m pip install -r requirements.txt

Yesterday, Cursor coughed up https://fdedictionary.com for me in PHP. You know what, it is like a dozen files and works awesome, no daemon needed, basically just a web server, some PHP files, and one SQLite file.

Today, Claude Code, the major hotness, made changes to a (different) toy app with less complexity, in node.js and you know what, it has ~49,000 files. And uses a 3rd party SaaS to host the db, and another 3rd party SaaS to do auth, and yet another 3rd party SaaS to import a repo and do deployments and hosting.

Simplicity has a place for some things. I love PHP+SQLite for tiny apps, and LLMs/agents are awesome at it.

I forgot, ChatGPT one-shotted the PHP app and Cursor put the finishing touches on. What a world. Long live PHP, is the point. Arrays and all.

That said, if you want to switch jobs (in general) you mostly need 1) the exact skills that that job requires, which are listed on job listings usually, so read those and b) probably you need to know someone who works there already to put your profile forward so you can get the foot in the door for an interview and rise above the 2000 other similarly qualified applicants. There are only so many industries and companies, it's not insurmountable to research.

What are you interested in doing at this "new job"? C, C++, python, OpenCL, GPU, DSP, and some docker again, but in more depth? Then you'd want to look for jobs that have those as requirements. Do you know people at any of the companies with those job listings? Is that realistic? Maybe, IDK, you can look and find out though.

Like I said, you are maybe luckier than you know.

I tell you my experience. I have >10 YOE. In a niche. And everyone whats to hire for the EXACT KNOWLEDGE THEY NEED not what some smart person has and can do once hired. So I cram for like 5 different domains for 5 different interviews, I tell you want, it's silly.

Today's hype/value cycle is in AI.

If you want a 30-day plan to upskill to deliver that to companies, see the post at https://news.ycombinator.com/item?id=48218162 and the tweet linked in the OP. The outline some kind of 30 day plan to learn those skills. If that sounds like something for you, maybe consider it.

If you just want generic depth in your favorite technical languages/topics, maybe start contributing to the open source projects in those languages that you like or make use of. Probably a pretty good way to get depth. But better to solve real problems for money, maybe. Because nothing ain't free anymore.

Honestly, best advice is this book IMO, "The Science of Getting Rich" by Wallace D. Wattles (by now old enough to be in the public domain):

  Read: https://www.gutenberg.org/files/59844/59844-h/59844-h.htm

  Listen (3 versions): https://librivox.org/author/848?primary_key=848

  1) always build on what you've built already in terms of expertise and experience and ability to add value, and

  2) always give people more value (at work) than they expect or think they deserve

One other advice, I find every morning it helps to think: how great it is this thing I'm working on today because it helps people who need _____ (that thing). It makes it more exciting. Maybe you don't need this one, but I thought I'd mention it.

https://apps.apple.com/us/app/forum-a-facebook-app/id6758308...

Posts:

https://www.engadget.com/2179165/meta-forum-groups-app/

https://www.threads.com/@mattnavarra/post/DYnoSuSCjXq

https://firethering.com/meta-forum-reddit-like-app-facebook-...

I don't expect uv to do anything fancy. I don't run module management (installation, upgrades) through uv commands. I don't much care what uv's syntax is. I let each tool do one thing.

- Tool 1: UV makes a venv for each project, with whatever Py version is suitable for that project, so that each project's dependencies do not collide with one another.

- Tool 2: Pip installs all the requirements for a given project, within the venv for that project. From a requirements.txt file. Which, as far as I am aware, pip commands and requirements fit more of what they author is looking for.

I don't think it's necessary to subject oneself to the things the author (articulately) complains about, e.g.: uv's command syntax for listing packages; uv's emitting unbounded requirements syntax; uv's command to upgrade modules

Then again, it's quite possible the author is managing modules in projects with more complex needs than I am.

Long-winded example:

  # Make an env for the project with appropriate Python and use it
  uv venv ~/.venvs/myprojpy312 --python 3.12
  source ~/.venvs/myprojpy312/bin/activate
  
  # Make sure pip exists and is up to date
  python -m ensurepip --upgrade
  python -m pip install --upgrade pip

  # Fill in requirements.txt in readable/meaningful syntax per needs
  $ cat requirements.txt
  requests>=2.31.0,<3.0.0
  black==24.4.2
  
  # Install the requirements initially (or again after changing requirements.txt) 
  python -m pip install -r requirements.txt
  
  # List outdated modules
  python -m pip list --outdated
  
  # Upgrade modules, respecting the constraints
  python -m pip install --upgrade -r requirements.txt

  $ cat ~/.config/uv/uv.toml
  exclude-newer = "7 days"
  # per https://news.ycombinator.com/item?id=47884491

Absolutely. To get 1/8 PB = 125 TB home library "easily":

We'll use 8 disks in the 125TB library. Between RAID 5 (1 disk lost OK to recover) vs RAID 6 (2 disks lost OK to recover), choose RAID 6 (our disks could fail at same time if of similar production or unlucky). RAID6 means 25% of space used for parity overhead, and 2-5% used for metadata/filesystem.

So looking for about 163TB. 163TB / 8 rounds to 21 TB. This pushes us above 16TB disks. Between 20TB and 22TB, choose 22Tb to feel safe.

Napkin math:

  Synology 8-bay DS: $1150 (Amazon price)

  8x 22TB Seagate 22TB external 3.5" = 8 x $390 = $3120 (also the #1 least expensive disk per TB for 3.5" external at https://diskprices.co currently)

But something cvan happen to that. Fall, fire, water, theft, party. We could lose everything.

So following 3-2-1 we'll have 3 copies, on 2 media, with one offsite.

Copy 2 can be same as first (RAID is for disk redundancy not backup -- we still have one copy only).

By now, 2x Synology 8-bays, plus 16x 22TB disks, puts us at $8540 for what we can keep at home.

But disks only really last about 5 years. They're getting kinda better, but in reality those disks can fail and should be replaced about every 5 years, some people get 10.

So every 5 years, we can want to shell ou;t about $8540. But wait, disks about doubled in the past year. Maybe it'll be $16,000 next time? Hard to say.

We still need a 3rd, off-site copy for 3-2-1. Recent reports indicated Backblaze silently lost data, some people exodused I believe. To where? IDK, but let's pick Amazon Glacier deep storage. At 125TB (just useful data), at $0.00099/GB/mo, that puts it at, over the same 5 years: $0.00099/GB/mo * 125000 GB * 12mo * 5yr = $7,425/5yr

(For the remote copy: can your ISP actually handle uploading 125TB? How long does that take to do once, even half? Is ISP transfer capped? Will 3rd party storage provider change prices or lose data? That's why we have 3 copies, maybe change providers when needed.

In any case, add it on 3-2-1 for 125TB would cost, at the easiest/cheapest: $4270 * 2 + $7,425 = 4270+7425 = $15,965, good for about 5 years.

So, every 5 years, spending $15,965.

At these volumes, are do even have ECC RAM? Are we scanning for and correcting errors with correct data when they occur? We don't want a hobby, we want an appliance, for this library, often especially if we work 99% of the time in tech and have life to live, quite likely.

Let's try another formula: on a shoestring and a hope, one could do it "cheap on RAID 5 (only 12.5% lost to parity and metadata/filesystem) and under-storage without 3-2-1" by going Synology 8-bay ($1150, Amazon) + 8 * 16TB (8 * $410 per https://diskprices.co = $3280) = $1150 + $3280 = $4430

---

  In grand summary, roughly every 5 years:

  Done "right": $15,965
  Done "cheap": $4,430 and only 112TB usable.

Averaged over 5 years (though it's not) these are:

  - Right: $15,965 / 5yr = about $3,200/yr (plus tax) for 125TB usable library
  - Cheap: $4,430 / 5yr = $886/yr (plus tax) for 112TB usable library

Maybe doable. But it's like owning another car in more ways than one (cost, maintenance/ongoing-care). Some individuals can swing it without even thinking. Most can't.

IMO, if the AI industry or any players would like us to become more computer centric, and make use of all the data that tech now lets us have, its constituents should do something (anything) to drive the cost of disks DOWN, not UP.

> why AI companies are in desperate need of them [FDEs]

> the only competitive edge is how and where you use it [intelligence]

> determining how and where companies use it [intelligence] becomes the most important role

> that is the role of a forward deployed engineer

> Businesses hire an Applied AI company...to help them get the most out of the technology.

> gives them access to a team that has already done large-scale AI transformations...make clients move much faster than their competitors

> FDE is a highly skilled engineer who can understand the customer's problems very deeply, write code into a code base they've potentially never seen before, and communicate the business impact to a non-technical decision maker to close the deal

> be on-site with a customer

> to see real efficiency gains, a company needs to be rebuilt around AI from the ground up...only possible through sitting with the customer and building custom agents that are engineered on company-specific data, with company-specific context

> three main parts of an Applied AI FDE's job: Audit, Evals, and Deployment

> Audit: ...onsite...mapping processes/workflows...example: two weeks with rev ops, one week with procurement, and a full month with finance...determining what should be automated vs what shouldn't...agents can create more problems than they solve...[elaboration]

> Evals: ...correct...thinking like a human would...[elaboration]

> Deployment: ...existing data layer...orchestrator to query...execution environment to test...debug...[elaboration]

> [how to break in]...build projects...explain every single component...30-day plan...

> FDE is the most in-demand role in tech right now