NHacker Next
  • new
  • past
  • show
  • ask
  • show
  • jobs
  • submit
▲How to Break into Security, Ptacek Edition (2012) (krebsonsecurity.com)
liquid_oil 20 hours ago [-]
Reading a post from 2012 like this it's shocking how quickly the ladder has been pulled up in all computer careers, but I suppose it's not as bad as the ladder pull from the dotcom bubble to 2012 where you could previously land a career "knowing HTML".
Aurornis 16 hours ago [-]
Framing it as the ladder being pulled up doesn’t feel right. The people in existing jobs aren’t deciding market conditions.

The market for juniors has had some wild fluctuations. During the ZIRP period even juniors with little experience were getting hired easily because there were more openings than applicants. Now the balance shifted the other way and there are (seemingly) more applicants than openings, so jobs go to experienced people first.

There are still ways for juniors to get in to the industry, but it takes work. The days of getting hired for having basic knowledge are gone for now, but like we saw in 2020 it could change rapidly again.

xyzzy123 19 hours ago [-]
Surely "knowing HTML != you're hired" after dotcom bust was not so much a ladder pull as a regression to mean from an unprecedented employment boom?

As for today, there was a shock from end of ZIRP and now AI.

On the AI side I guess one could read it as a ladder pull but it feels more structural to me? The entire field is in an existential crisis while we try to figure out what our jobs are now, what hiring criteria we should use, how work should be organised and what products might even be valuable anymore.

For boutique security specifically I can't say if it was difficult or easy to get hired in 2012, sort of both? The hardest part was figuring out where the ladder even was. It was hidden round the back of the building, next to the loading docks where people smoke. The hiring path was quietly "game recognise game", i.e, you learn & hack stuff in a safe way and people already doing that recognise that you can do the job. It felt a lot more like an apprenticeship system than anything formal. Certificates never mattered - that's not what anyone was looking for (by the time someone can certify you to do it, it's not really hacking anymore). The weird part is that in 2026 there are some clearly marked bright orange ladders to get into the field but because they are clearly marked there is now a queue for them.

dhosek 18 hours ago [-]
In 2006–7, knowing HTML and CSS could get you $30/hour still which was years after the dotcom collapse.
leesalminen 17 hours ago [-]
Fond memories! Circa 2006 I was making $20/hr slinging <table> elements all day. Was more fun than working at McDonald’s and paid 3x to boot.
sam_lowry_ 10 hours ago [-]
Front-end developers around me do not know HTML and CSS. And I am not kidding, they jump straight from the uni into Angular courses, then into the job where they just hurdle agents around Angular directives.
sorentwo 20 hours ago [-]
Marking this as 2012 would help. It’s a fun read in a nostalgic, pre-ai kind of way, where emphasizing that developers should use bcrypt wasn’t standard practice.
michael-bey 6 hours ago [-]
It's still pretty good advice. I would replace "learning how to program" with "learning how to critically evaluate LLM output".

Like almost everything these days, it's K-shaped. High agency people who use LLMs as a tutor will be able to more quickly break into security. The script kiddies will still script kiddy, except instead of running a random Python script it's doing it uncritically from Claude / Cursor / Codex.

As someone who had to break into security myself around this time, I would have loved an LLM instead of trying to cobble together understanding from forum posts and PDFs.

bryant 20 hours ago [-]
tptacek may disagree, but this needs an update considering how much has changed both with AI and the economies of scale AI brings to the discipline.

I'm anticipating "the fundamentals are crucial," and as much as I want to agree, I think practical realities show us otherwise. We're at a point where all someone needs to know is how to AI-program-manage and how to glean enough about the discipline (or any discipline at this point) from AI to be dangerous enough in the field. To be a builder, breaker, or defender wearing any color hat of their choosing... that's it, really.

flanz_10 20 hours ago [-]
> but this needs an update

it's a decade+ old document. i dont think it needs anything, other than a notice of its publication date.

at this point, much more than an update would be required to be up to date

wavemode 17 hours ago [-]
Sorry - what specific part of this article is now wrong because of AI?
subroutine 20 hours ago [-]
“tptacek may disagree” - famously. The only uncertainty is whether he’s seen the thread yet.
smashandgrab 19 hours ago [-]
I hire for incident response roles. Honestly, despite how much has changed this is still the way to do it for us. You need to be versed in many things and more importantly, interested enough to not feel entirely stressed out when things hit the fan.

As for AI it’s been great for things like log review but those were already pretty low skill requirements. We haven’t had much luck past that.

j-scott 20 hours ago [-]
(2012)?
elevation 20 hours ago [-]
Yes -- I was keen to hear what tptacek had to say about breaking into the industry in the age of LLMs and 0-days being discovered by the dozen for pennies.
tptacek 20 hours ago [-]
At this point, who knows?
18 hours ago [-]
smashandgrab 19 hours ago [-]
What you said combined with you being the author and you being at the bottom of the comments (at the time I wrote this) is confusing and hilarious.
BiscuitBadger 18 hours ago [-]
Encouraging words!

Seems these days employers want devops, SDE, product designers, and security and they want it all for $60/hr in office.

Arch-TK 18 hours ago [-]
Security pays well, but software engineering still pays better. Even significantly with the right luck and skills.
dzonga 20 hours ago [-]
it would be interesting to know what tptacek's thoughts are in the age of 'agents' violating vulnerabilities willy-nilly ? if his thoughts of getting into application security still apply.

what the incentives are since hacking seems to be no longer a crime given how many hacking incidents have gone unpunished.

uptrails 20 hours ago [-]
Now would be a very weird time to write any kind of an authoritatively-sounding "here's how to break into software engineering" or "here's how to break into infosec".

On some level, nothing has changed - you still get noticed and hired the same way as before, the companies are still pretty much the same - but I just wouldn't have the confidence to tell anyone "hey kid, this is how it's going to look three years from now". It's not that I think these roles are going away, but there's a good chance they're going to be a lot less interesting (and not as well-paying) when it's all said and done.

david_shaw 19 hours ago [-]
I'm often asked how to turn a security-related interest or hobby into a job. For the longest time, I'd say pretty much the same thing written here: get a book, do some research, build some projects, talk to people in industry. I've been going to DEF CON for about 25 years now, and I've had this conversation countless times.

But in 2026, tptacek says it best in the comments here: At this point, who knows?

Are we still telling kids to learn programming languages to become software developers? Perhaps system design is a more reasonable skill set.

That said, because this is the time to say it: security is still important, and it's not -- and shouldn't be -- fully automated.

The toolchain is changing just like it always has. I remember when interpreted languages replaced compiled ones, and people said that interpreters solved security. Then people went from writing their own codebases to implementing frameworks, and people said it solved security.

Barracuda Networks posters at every airport stated "Security Solved."

As long as people take risks, security is important. Now we just need to focus on designing resilient systems instead of finding bugs and vulns on the code-level. LLMs can do that part for us.

jamesy0ung 19 hours ago [-]
> I'd say pretty much the same thing written here: get a book, do some research, build some projects, talk to people in industry. I've been going to DEF CON for about 25 years now, and I've had this conversation countless times.

This advice is still true in 2026.

19 hours ago [-]
chias 11 hours ago [-]
Does anyone want my spot?
cyberax 14 hours ago [-]
What is even security, these days?

Algorithms are a solved problem. Post-quantum encryption was the last bit of excitement for a while. I guess ZKPs and FHE are still available, but they are really niche.

It's everything else that fills me with dread: we're using oodles of software written in C/C++, so we're one missing array boundary check away from the global economy collapsing.

Then there's phishing and social engineering. If one of our developers gets compromised, it's trivial to push malicious code into our repo. And from there it's an easy escalation into our customers' billing, private information, etc. We store the bare minimum of data possible, but it's still a lot.

ronsor 14 hours ago [-]
FHE is still really cool, if the performance issues can be solved.
cyberax 14 hours ago [-]
It's cool, along with ZKPs but I struggle to find a useful niche for them. Age/ID verification is cool and all, but this is apparently it?

(And I'm ignoring all the blockchain nonsense on purpose)

ronsor 10 hours ago [-]
FHE is good whenever you want to perform a computation remotely without losing privacy, so anything from AI inference to search indexing.
silentsobs 15 hours ago [-]
[flagged]
Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact
Rendered at 18:58:36 GMT+0000 (Coordinated Universal Time) with Vercel.