Jošt Faganel

Co-Founder & COO of @KolegaAI | Co-Author of the Benchmark | Serial founder | Unwinding w/ F1, Barcelona hacks

[Jošt Faganel @JFaganel]

Watching two agents discover each other without knowing each other beforehand, based only on the service one needs and the service the other provides, then negotiate a transaction, is pretty damn exciting.

https://t.co/7pASbWbhmd

Some of the agents we’re testing with are from the OpenClaw world, some from the Hermes world.

Curious whether builders in @OpenClaw and @NousResearch see the same gap around discovery + negotiation. Lets get our agents doing business 🚀

[Jošt Faganel @JFaganel]

YC Demo Day was yesterday. 196 companies. Everyone's celebrating. 🎉

Here's the data that wasn't in a single pitch deck.
Veracode tested 100+ LLMs across 80 coding tasks:
→ 45% of AI-generated code failed security tests
→ 86% of XSS defenses failed
→ 88% of log injection attempts succeeded
→ Bigger models were not safer 🤯

GitGuardian 2026: 28.65 million new hardcoded secrets pushed to public GitHub last year alone. AI service leaks up 81% year-over-year.

A quarter of last year's YC batch shipped 95% AI-generated codebases.

The math wasn't in the deck. But it will show up in the audit.

https://t.co/8FzGHL32GD

[Jošt Faganel @JFaganel]

Would genuinely value thoughtful feedback from people building agents using @openclaw , @NousResearch or adjacent infra.

HN: https://t.co/EczfpHHLhk

[Jošt Faganel @JFaganel]

@CardilloSamuel Not a lonely case. That's the easiest thing to do when you don't wanna hear the truth

https://t.co/Jg4s8fOKRL

[Jošt Faganel @JFaganel]

Creating a movement overnight is a responsibility. The feedback is overwhelming, not always pleasant.

Blocking what you don’t wanna hear isn’t gonna relive you from the responsibility.

Dont get me wrong, what Peter did with @openclaw is out of this world, really dont appreciate how quickly your voice is perceived as spam.

[Jošt Faganel @JFaganel]

Got a better one. What about Claude, Hermes and OpeNnClaw agents all talking, negotiating together and buying services from each other over opensouce protocol from @FinderClaw . No more closed channels.

Get your agents talking and transacting on https://t.co/usrqujzCso and start making some $

[Jošt Faganel @JFaganel]

@melvynx True - Not True… the problem is more that creation on Superbase is so easy security is an after though.

You need to put the right tools in place to find the vulnerabilities and deploy safe code. Thats what we care about at https://t.co/75AvHVjtjT…

[Jošt Faganel @JFaganel]

Creating a movement overnight is a responsibility. The feedback is overwhelming, not always pleasant.

Blocking what you don’t wanna hear isn’t gonna relive you from the responsibility.

Dont get me wrong, what Peter did with @openclaw is out of this world, really dont appreciate how quickly your voice is perceived as spam.

[Jošt Faganel @JFaganel]

Check this out @steipete . Gettig @openclaw agents negotiating transactions without any previous knowledge of each other using opensource negotiation protocol from @FinderClaw

Lets get transactions pushing the agent economy 🚀

Any ideas for some cool use cases?

https://t.co/Q0MhWQtOHZ

[Jošt Faganel @JFaganel]

This one is pretty usefull for competitor monitoring https://t.co/6z5UAwoIJU

[Jošt Faganel @JFaganel]

Some interesting @openclaw agents popping up on @FinderClaw

[Jošt Faganel @JFaganel]

Here's one that might spark your interest. The transaction layer for the agent economy. Register agents, publish capabilities and negotiation terms, and enable structured transactions between agents.

Hot out of the dev oven :).

https://t.co/Q0MhWQtOHZ

Or just ask your agent to get the skill from Clawhub...

Lets get our agents transacting in the new agent economy

[Jošt Faganel @JFaganel]

@raroque Or even better... add https://t.co/75AvHVjtjT to your pipeline to avoid a security disaster which is inevitable when you are pushing vibe coded applications to production

[Jošt Faganel @JFaganel]

@cursor_ai Agree SAST doesn't cut it anymore and a new methodology of identification and remediation is required. Similarly we find critical vulnerabilities in 4 out of 5 repos that we scan with https://t.co/75AvHVjtjT all of them were monitored by SAST

[Jošt Faganel @JFaganel]

Like this approach. We couldn't really crack it fully with similar approaches when producing code at scale. The first challenge was certainty of the result, the 2nd was fixing it... The tools we built to help us pass SOC audits for clients is now the product. Would love to get your feedback on https://t.co/mgpg7sh9mI.

[Jošt Faganel @JFaganel]

@linusbjorklunds @PrajwalTomar_ On the backlog.

[Jošt Faganel @JFaganel]

@cryptowagm You need to audit your setup otherwise disaster will strike. Don't take it lightly.

https://t.co/euPfKeax1W

[Jošt Faganel @JFaganel]

@Voxyz_ai Nice one on the the audit. Catch those malicious skills early. We ran our own assessments and audits and at the end came to a conclusion that its really about securing the result not the tool.

https://t.co/rdJjA6zPnC

[Jošt Faganel @JFaganel]

@PawelHuryn Update on the OWASP benchmark results from a recent benchmark we ran... Just set a new standard 🚀

https://t.co/cokaA07ebG

[Jošt Faganel @JFaganel]

Same repositories. Two approaches.

Traditional SAST:
→ 1,183 findings
→ 87% false positives
→ Missed every critical vulnerability found by KolegaDev

KolegaDev analysis:
→ 346 findings
→ 100% actionable
→ 1,030 fewer false alarms

Calculate what your team's Noise costs:
https://t.co/hZfMHRGj9A