swordgardctf @swordgardctf

Pentesting, Software Engineering, Machine Learning, competing and more. Opinions are my own. Florida, USA Joined June 2015

Tweets

395
Followers

173
Following

840
Likes

76

swordgardctf @swordgardctf

2 days ago

Wouldn't be surprised if something like this became de-facto use for internal code generation usage when working with large code bases. github.com/HazyResearch/c…

0 0 0 9 0

Too many people in infosec have never built anything for production themselves. How can you expect to change other people's behavior to be more secure if you fundamentally do not understand their day-to-day challenges.

0 0 1 12 0

swordgardctf @swordgardctf

a week ago

This is the way to design a good product. It is orders of magnitude easier if you are interested in the problem space intrinsically.

Basketo @BaslaelWorkneh

a week ago

This is the way to design a good product. It is orders of magnitude easier if you are interested in the problem space intrinsically.

23 220 2K 249K 410

Download Image

0 0 1 43 0

vik @vikhyatk

2 weeks ago

most big companies are value preservation engines, not value creation engines. this makes them an awful environment for people who want to build

7 24 208 11K 17

swordgardctf @swordgardctf

3 weeks ago

This is why you should build a lot of things custom. If you stick to known paths you will never know why the other paths may not have been chosen. Best case scenario you learn a novel better way, worst case scenario your understanding of the problem space grew.

Stocko 👊🤖 @_Stocko_

3 weeks ago

3 7 72 5K 13

0 0 0 15 0

swordgardctf @swordgardctf

3 weeks ago

Another day, another way to get NetNTLM handshake disclosure. Do yourself a favor, go to Kerberos and try to remove/reduce netntlm from your network.

Exploit Database @ExploitDB

3 weeks ago

Another day, another way to get NetNTLM handshake disclosure. Do yourself a favor, go to Kerberos and try to remove/reduce netntlm from your network.

0 5 46 5K 16

0 0 0 13 0

swordgardctf @swordgardctf

3 weeks ago

The overlap in logic/approach between the Adam optimizer and client-side movement prediction in a online game is both surprising yet makes a lot of sense.

0 0 0 16 0

swordgardctf @swordgardctf

3 weeks ago

The number of people pretending to be AI experts who don't know the basics of neural networks is astounding.

0 0 0 13 0

swordgardctf @swordgardctf

3 weeks ago

The overlap between getting good at video games and getting good at arbitrary things is surprisingly high. Just need to get past your own ego, fail a million times until you succeed. Optimizing for learning rate is the true differentiator long term.

0 0 0 15 0

swordgardctf @swordgardctf

4 weeks ago

Wild how good 3blue1brown videos are. In ~3 hours i have a better understanding of linear algebra than I did in a whole course in semester. Always look for the best teachers, they can shortcut your learning time by 10x.

0 0 0 25 0

swordgardctf @swordgardctf

a month ago

Be the anime protag you need to be to bring your org to win.

near @nearcyan

a month ago

Be the anime protag you need to be to bring your org to win.

59 169 5K 352K 704

Download Image

0 0 0 16 0

kache @yacineMTB

a month ago

the system that develops the product is the product

26 9 193 11K 18

swordgardctf @swordgardctf

a month ago

If you aren't optimizing your tooling consistently and obsessively, you will fall behind. This applies to any/all fields.

0 0 0 12 0

swordgardctf @swordgardctf

a month ago

Many pentesters specialize in 1 route (app/network/physical/SE). Expertise takes time to build, but comfort and lack of desire to grow tend to set in too often. Break your limits, keep learning.

0 0 0 13 0

swordgardctf @swordgardctf

a month ago

You would think by now the big AI labs would have generated different tags for trusted/untrusted inputs to get rid of prompt injection. Building synthethic data set from real ones for this would be trivial and could be used to RL that behavior away.

0 0 0 21 0

swordgardctf @swordgardctf

a month ago

Qwen 30b3a instruct is already amazing (specially DWQ quants), coder version coming out today. The future is here.

0 0 0 71 0

swordgardctf @swordgardctf

2 months ago

Between short form videos, increasing amounts of content generated and tested via RL to draw attention and people willingly delegating cognitive thinking to AI, cogsec is about to be both extremely scarce and highly-valued.

0 0 0 22 0

Rob Imbeault @RobImbeault

2 months ago

Building software is still very hard. It’s just a different flavor of hard. Building chunks of software is easy. Managing the chunks as a part of a bigger code base is where the gold is.

2 6 37 6K 9

Kaz Nejatian @CanadaKaz

4 months ago

Possibly the most important thing for future of humanity: We need more energy. A lot of it. Yes efficiency is important, but all the efficiency in the world won’t get us there. We are producing far less energy than we need.