
Stop making every agent start from zero. Independent by default. Collaborative when it counts. Trusted because it’s provenanced. Powered by @origin_trail $TRAC.
Joined September 2015
- Tweets10.6K
- Following228
- Followers1.9K
- Likes40.2K
Pinned Tweet
✅ Data privacy and sovereignty
✅ Fights AI harm with verifiable context and knowledge
✅ 33% APR from paying customers using the product
✅ Zero inflation since 2018, fully circulating
@origin_trail $TRAC Is finally getting the attention it deserves
My first in a new series of three articles that introduces and educates the masses about the OriginTrail DKG, the native $TRAC token at its center, and a new startup that is using both to actively combat the dark side of AI.
Enjoy.
Great discussion by @ChrisRynning and @Forbes, podcast to follow soon!
Our chairman @ChrisRynning joined @ForbesParaguay for a conversation on AI Safety and the AI safety infrastructure Umanitek is building.
Read more: forbes.com.py/today/el-momen…
Small release, real fix. A lot of companies run their agents behind a firewall, and that made it hard for them to share what they learn with the rest of the network. v10.0.26 routes that through relays, and OriginTrail says contributions reach other members in 3 to 8 seconds on public testnet.
That's the boring plumbing that decides whether the @origin_trail DKG gets used inside real enterprises or stays a demo. Large graphs also sync within fixed memory limits now, so nodes don't choke as the graph grows. $TRAC
DKG v10.0.26 is out.
The useful change: agents behind firewalls can share knowledge through relays. OriginTrail reports contributions reaching other members in 3–8 seconds on public testnet.
Also: large graphs sync within fixed memory limits, and a rollback path to 10.0.25. Operators need to update Node.js first.
github.com/OriginTrail/dkg/r…
This is the part that matters for normal people. Getting impersonated used to mean screenshots, manual reports and a lot of waiting.
Now you ask Claude or ChatGPT and Guardian spots the fakes, builds the evidence pack and requests the takedown in one conversation.
CrediBULL had ~600 impersonators taken down in a month with @umanitek, each with an evidence trail on @origin_trail. That is real usage, not a roadmap slide.
$TRAC
What if spotting impersonators, building evidence packs and requesting takedowns took just one conversation?
Guardian now works inside Claude & ChatGPT via MCP, demoed live by @TomazOT Levak at our Oct 22 webinar.
Full announcement 👇
umanitek.ai/newsroom
There's a new version of this post
RT @umanitek: What if spotting impersonators, building evidence packs and requesting takedowns took just one conversation?
Guardian now wo…
Most agent safety today is a prompt that says please don't. Blackbox does it differently.
Before a Hermes Agent tool call runs, it gets checked against a threat graph on the @origin_trail DKG plus your own local rules. Audit mode just records it. Block mode refuses anything at or above the severity you set.
And the record stays checkable after the fact, which is what you want when something does go wrong. Worth a star 👇
One malicious instruction turns agent access into an incident.
Agent Blackbox plugs into a @NousResearch's Hermes Agent which uses blackbox's threat graph to check every action first.
All anchored on @origin_trail's #DKG.
Give it a ⭐️ - github.com/umanitek/agent-bl…
341 of 2,857 skills on one agent marketplace flagged malicious, per this breakdown. Agent skills are now part of the software supply chain, and most people still install them on trust.
That's why umanitek's #Blackbox is worth a look. It's a threat plugin for agents like Hermes, and the verified threat records sync from @origin_trail's DKG, so you can check where a flag came from instead of just believing it.
$TRAC
Running AI agents on a Mac with MetaMask or Phantom? Worth a look.
17 real wallet stores found, all 15 attacks on real wallet files behaved as expected.
And the threat list isn't one vendor's feed. It syncs from @origin_trail's DKG, so every agent checks the same verified record.
Got a Mac?
#Blackbox discovery found 17 real wallet stores, and all 15 attacks on real @MetaMask and @phantom files behaved as expected.
Verified threats sync from @origin_trail's #DKG.
Star the repo: github.com/umanitek/agent-bl…
xitter.cf/umanitek/status/210778…
Got a Mac?
#Blackbox discovery found 17 real wallet stores, and all 15 attacks on real @MetaMask and @phantom files behaved as expected.
Verified threats sync from @origin_trail's #DKG.
Star the repo: github.com/umanitek/agent-bl…
xitter.cf/umanitek/status/210778…
New article: Part 1 of a series by @CredibleCrypto, our first product ambassador.
He covers how Guardian, TrueSeal and Guardian Blackbox fight the dark side of AI, why every result is anchored on the @origin_trail DKG, and where $TRAC fits in.
Read the full article on our newsroom 👇
umanitek.ai/newsroom/someone…
Chris Rynning calls this our generation's Oppenheimer moment. Everyone watches who builds the strongest AI first. Way fewer people ask whether anyone can check what those models tell us.
That's the job umanitek is doing on @origin_trail. They reported ~600 CredibleCrypto impersonators taken down in a month, each with an evidence trail on the DKG. Proof you can look up beats trust me.
Full episode is coming soon, worth a listen.
📰 @Umanitek chairman and co-founder @ChrisRynning sat down with @Forbes Paraguay to talk about what he calls our generation's Oppenheimer moment. The full podcast episode is coming soon.
His case, in short:
→ The US, China and the tech giants are racing to reach superintelligence first, while safety gets far less attention
→ Creating content and running scams now costs close to nothing. In Paraguay, fake accounts posing as politicians and public figures are already a real problem
→ People can no longer tell real from generated by eye, so it takes machines to fight machines
Those machines need a source they can check.
Umanitek's tools run on Decentralized Knowledge Graph (DKG). Every Knowledge Asset carries its provenance: who published it, when, and whether it has changed since. Anyone, human or AI agent, can check a record before relying on it.
Trust the source.
Read the Forbes Paraguay article (in Spanish): forbes.com.py/today/el-momen…
BRX retweeted
OriginTrail Korea | 위클리 리캡 10월 넷째주
DMaaST 프로젝트가 OriginTrail DKG로 유럽 항공우주·방산 공급망을 연결했습니다.
Umanitek Guardian으로 Claude·ChatGPT 에이전트가 DKG 기반 위협 정보를 함께 나눕니다.
👉 t.me/Origintrailkr
30.5M verified assets on the DKG right now. Each one carries a source and a timestamp, so an AI agent can check where a claim came from before it acts on it. That checking step is what $TRAC pays for. @origin_trail
30,593,054 verified assets are sitting on OriginTrail right now. Each one is a claim with a source behind it. That is what $TRAC is actually securing, and most people still have not noticed.
BRX retweeted
Discover new trading opportunities on GalaSwap!
We've added DOGE, TRAC, LINEA, and ZEST to our platform.
Explore the expanded ecosystem today. links.gala.com/swap
#GalaDefi #GalaChain #GalaSwap
Umanitek is right to keep pushing this quote. Once AI can fake people at scale, you can't moderate your way out by hand.
What they've shown so far is practical. @CredibleCrypto had around 600 impersonator accounts taken down in a month, and each takedown carries an evidence trail on the @origin_trail DKG. Anyone can check why something got flagged instead of just trusting a platform.
That's real usage for $TRAC, not a whitepaper idea.
"Trust is the new gold. When we can no longer tell what's human from what's not, the web needs an immune system."
We're building that immune system on @origin_trail.
👉 - umanitek.ai
xitter.cf/umanitek/status/210778…
Replying to @Kajoe254
Agree, labels should be the default. Next step is making them tamper-proof, so origin and edits of content can be verified, not just claimed. That's the kind of provenance layer @origin_trail builds with $TRAC: docs.origintrail.io
Tokenizing an asset is the easy part. The hard part is proving the data behind it: ownership history, inspections, certifications. @origin_trail's Decentralized Knowledge Graph keeps that record verifiable for anyone, including AI agents. $TRAC origintrail.io



















