Given the state of AI we must internalize: Everything hackable will get hacked.
Two forces are at work:
βͺοΈ Near-frontier open weight models hack without refusal
βͺοΈ Frontier models help you defend yourself
How we're approaching this new reality β
We compressed the onboarding to deepsec from
1. Run init
2. Copy paste a prompt into clanker
3. Run scan
4. Make a judgement call if matches are good enough
4.1. If not, tell clanker to make custom matchers
4.2. GOTO 3
5. Run processing
To just running init. If deepsec felt a
Your first deepsec security review is now one command.
πππ‘ πππππππ ππππ
Start the review, then come back to security findings across all of your existing code.
vercel.com/changelog/simpβ¦