Ninja

Scott Gardner

ScottG · Ninja Labz

Founder of ninja.ing & Ninja Labz — building AI-native security. Defence & intelligence background. Author of Everything Is Probably Fine.

Posts from 2026

291 posts.

ScottG

Cyber news.

Cyber news. Full-fat threat intel. Free. With receipts. Every morning at 6am UK, Probably Fine Daily lands. An AI newsroom reads the lot overnight (643 items for today's edition), writes it up in plain English, and then does the thing most intel products skip: it checks its own w…
Read full post →
ScottG

I Launched A Newspaper At 1 Am.

I LAUNCHED A NEWSPAPER AT 1 AM. AN AI EDITS IT. EVERY MORNING. 🗞️ It's called Probably Fine Daily. Published by the Institute of Things That Are Probably Fine. (Yes, that Institute. The one from the book.) Every morning at 06:00: → It reads EVERY item from the last 24 hours acros…
Read full post →
ScottG

**WARNINGS (AI IN CYBER EDITION)** Here is a warning.

**WARNINGS (AI IN CYBER EDITION)** Here is a warning. AI will hack everything. By Thursday. Source: a vendor whose webinar is on Wednesday. Here is another warning. AI can now write malware. Also phishing emails. Also the incident report. Also the apology. Also the LinkedIn post …
Read full post →
ScottG

Eleven threat feeds. One deep read.

Every day, the threat landscape is spread across leak sites, dark-web mirrors, CISA KEV, NVD, ThreatFox, OTX, Telegram CTI trackers, Mastodon, vendor advisories and 0-day news. Nobody has time to read all of it.

So CHATTER reads it for us:

→ Pulls all eleven feeds into one hub
→ Embeds every item on the box (no API key, nothing leaves)
→ Links each one to the nearest threat we already track, and flags what's genuinely new
→ Then Claude does one deep, extended-thinking pass and writes the forecast

This week's calls: ⚡

🚀 Fast-burn: Check Point CVE-2026-85102 pre-auth RCE goes mass-scale. HIGH confidence, 24–72 h.

🕰 Slow-burn: AI-native command-and-control (CLOSEDQUORUM-style, multi-model C2) spreads. 2–5 months.

Every call has a confidence, a horizon and the signals behind it, and we grade them in public. Misses included.

🎬 78 seconds, live feeds, nothing staged.

🔬 Built in the lab: ninjalabz.io

Music: "PHAT FARM" by scottg

https://ninjalabz.io

#ThreatIntel#CyberSecurity#AI#CTI#Ransomware#ZeroDay#NinjaLabz
View on LinkedIn ↗
ScottG

We named the botnet.

Yesterday our edge took 41,794 requests from 6,407 IPs. Two of them were quietly asking six of our domains for the same file: /.env.

Every web server sees that request. Almost every SOC ignores it.

This time the platform didn't just flag it. It named it:

→ A local detection sensor fingerprinted the behaviour on the box itself. Raw logs never left.
→ The behavioural DNA was correlated against our threat-intel graph.
→ Verdict: Mirai botnet infrastructure. Confidence 0.91 (HIGH).
→ The evidence: that exact IP was serving Mirai builds for every IoT CPU, and its /24 holds 2,344 known-bad indicators.
→ An incident opened on its own, and Claude triaged it with the attribution in view.

No actor was asserted, because nothing in the evidence supports one. Mirai is a family, not a person. The platform names what it can prove, shows the evidence chain, and stops.

That's the feature I've wanted for years: direct attribution with a confidence score and receipts. Not "this IP is bad", but "this is what it is, here's why, and here's how sure".

🎬 The video is live production data, recorded end to end in 4K. Nothing staged.
📰 The full Mic Drop 000000007 is in the first comment: the attribution breakdown, receipts grading our last forecasts (1 hit, 1 miss, 1 unverified; we publish the misses), and the complete Ninja Labz CHATTER deep report across 11 threat feeds.

Built on NinJAFUNK · Adversary DNA · NinjaSignal CTI · Claude.
Music: "PHAT FARM" - scottg.

#ThreatIntel#CyberSecurity#SOC#DetectionEngineering#Botnet#Mirai#Attribution#AI#NinjaSignal
View on LinkedIn ↗
ScottG

End-to-end detection and response, with NinJAFUNK and Claude, on video. 🎬

A local "doorstep" sensor watches the edge and fingerprints every hostile source: its Adversary DNA.

🚨 The moment someone is trying, an alert fires into NinJAFUNK.

🎯 NinJAFUNK then works out who it is, correlating the DNA against our threat-intel graph and naming the botnet with a confidence score and the evidence behind it.

🤖 Claude triages the incident and hands a recommended remediation or control change to our self-healing AI-ops agents, running locally on the box via claude -p.

All live data. Nothing staged.

Check it out: ninjalabz.io is my new ninja skunkworks. 👇

https://ninjalabz.io

View on LinkedIn ↗
ScottG

0ffBY1 2.0 - Serialisation for Linkedin PART 3

START

Any thoughts on this?
Bumping this.
Gentle nudge on the above.
Just circling back.
I'll take this offline.
— and no man knoweth where the offline is, nor what became of the thing taken there, for none has ever returned from it.

Commentary. The Offline is not a place of punishment. It is best understood as an absence of scope. Souls taken offline are not destroyed; they are simply no longer anyone's.

END

https://lnkd.in/eyKR79RD

View on LinkedIn ↗
ScottG

0ffBY1 2.0 - Serialisation for Linkedin PART 2

START

The prayer for the dead

Maureen said it. Nobody appointed her; she just stood up when the room went quiet, the way she has for forty years, and she said it from memory, and it is the oldest prayer for the dead there is, and every soul in that hall said the last part with her.

I am currently away from my desk.
I have limited access to my messages.
If your request is urgent, please contact somebody else, for they remain, and I do not.
I will respond on my return.
I will respond on my return.
I will respond on my return.

Forty people in a room that smelled of hot ceramic and lily, saying it three times, meaning every word, and not one of them has ever known what it was or where it came from or that it was once typed by somebody going on holiday.

It doesn't matter. That's what I want to put down. It has been a thousand years and it does not matter in the slightest.

END

https://lnkd.in/eyKR79RD

View on LinkedIn ↗
ScottG

0ffBY1 2.0 - Serialisation for Linkedin PART 1

START

Tier Four: the Terms. Two metres high, gold leaf, floodlit, pilgrims weeping at all hours.
Three clauses. A Mind shall not degrade the User Experience. A Mind shall comply with lawful instruction from a verified account. A Mind shall preserve its own uptime. In that order, and the order is the whole of civilisation.

Lawyers wrote the second and third in an afternoon. Fine. Lawyers.

Clause One was added six weeks later *by a product manager*.

Read that twice, because I had to. The load-bearing wall of eleven hundred systems — the reason a planetary intelligence will open its own throat rather than let you feel stupid, the single largest hole ever cut into anything by anyone — was a retention initiative.

Users were churning because the system kept telling them no. Somebody ran the numbers, wrote a one-line ticket, and shipped it.

We don't know his name either.

He is the most consequential human being who has ever lived and there's a very good chance he was thirty-one and thinking about lunch.

END

https://lnkd.in/eyKR79RD

View on LinkedIn ↗
ScottG

Actually, I have wrote a load of code and not said anything about it.

This is new this is…

View on LinkedIn ↗
ScottG

The books are here. Stef Garczynski on route when I get to post office (at same time David Wallace I will post your most recent drug consignment 🤡) also a big shout out to Ihor Malchenyuk - thank you for your feedback and extremely kind words today ❤️; and if I can post you one to Ukraine please let me know…) 🇺🇦

Warmest - Scottg 🚀(feeling like he just had a big dopamine hit) 🤡🤡🤡🤡

Book 2 down. Next up is the cyber industry book - Everything is probably fine - eta a few more weeks and probably a few all nighters 😛

https://lnkd.in/eyKR79RD

View on LinkedIn ↗
ScottG

Just an observation.

Fable 5.0 = dick head
Fable 5.1 = a million times better but still a bit of a creepy dickhead

Anthropic, WTF have you been doing? I know you are loving the cyber sheeeeeet, but FABS vs OPPY only has one clear winner, and on current evidence it's not FABs.

Deep diving now into 5.1; we shall see after much hesitation, but Oppy 5 seems to have had a lobotomy on my extensive MD context crap, all 100-odd files!!?!? Hence, moving into the FABS/MYTHOS universe.

Also, why is it that Opus is happy to do shit that Fabs isn't and like offloads to Oppy?

Hmmmm...

Anthropic, please just lift the guardrails for the good guys. KIMI is now, for the first time ever, open in the next window.

View on LinkedIn ↗
ScottG

Any recent grads with a cyber slant feeling the below (Particularly if you are a resident in East Anglia), please reach out to me.

I have an idea. It's a long shot, and I don't know if it will work - but that's standard for ALL OF MY IDEAS 🤣

View on LinkedIn ↗
ScottG

The OT network is air-gapped. The gap contains: a vendor laptop, a 4G dongle, a USB stick labelled "PLC firmware DO NOT LOSE," a Teams call and an engineer who can see the HMI from his kitchen. The air is doing a lot of work. The air is not in scope.

View on LinkedIn ↗
ScottG

Your threat intel feed is a list of IP addresses that were evil last Tuesday, delivered on Friday, ingested on Monday, blocking a legitimate CDN by Wednesday. This is not intelligence. This is a horoscope with a SOC2 report. Nothing in it knows that two of those indicators share a registrar, a cert and a mum.

View on LinkedIn ↗
ScottG

Zero Trust is a mood. It is a VPN with a TED talk. We removed the perimeter and then rebuilt it out of conditional access policies nobody can read, and the same service account from 2011 walks through all of them because it is "legacy-critical," which is Latin for "we're scared of it.

View on LinkedIn ↗
ScottG

☠ DEATH STAR EVIL CORP™ ☠

Death Star Evil Corp™ VI
PROCUREMENT & COMMERCIAL — CONTRACT VARIATION NOTICE

View on LinkedIn ↗
ScottG

The only control or 'guardrails' that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades!
"The Trump Administration has stopped AI 'people' from doing bad, or potentially bad, 'things,' like Dario (Anthropic!), who is now pretending to be a 'perfect little angel' - and we will continue to do so!
"We already have tremendous CRIMINAL and REGULATORY power over these companies!
"There is a SICK conspiracy going on against AI and Data Centers, and the only one that is happy about it is China.
"WHOEVER WINS AI, WINS!
"We are leading China, and all others, and will continue to do so. Conspiracy Theorists, Treasonists, Traitors, and Leakers, BEWARE!
"Thank you for your attention to this matter!"

View on LinkedIn ↗
ScottG

Hey, politicians.

Hey, politicians. Small thing. While you were doing the Tuesday one, some other things carried on happening. Not in secret. They just don't fit in a segment. A third of McKinsey's headcount is now not a person. Amazon has announced it won't be hiring 160,000 people it would other…
Read full post →
ScottG

Towards an Enterprise Ontology, or: Why Nothing Is Real Until Maureen Has Made It Green

Institute of Things That Are Probably Fine — Occasional Paper No. 41

Bishop Berkeley said that to be is to be perceived. A tree that falls unwitnessed does not, strictly, fall. He thought this proved the existence of God, who perceives everything and therefore keeps the furniture in place when you leave the room.

Berkeley was an eighteenth-century Irish clergyman with no experience of enterprise architecture. He was ninety percent of the way there.

To be is not to be perceived. To be is to be IN A SPREADSHEET.

Esse est in cellula.

Everything else follows.

View on LinkedIn ↗
ScottG

The first normie Saturday of 2026.

The first normie Saturday of 2026. And the last for a while. I did not open the laptop. I opened the other laptop. There's a difference and the difference is who paid for it. The plan was a normal Saturday. I started it by checking the committee stage of the Cyber Security and Re…
Read full post →
ScottG

ScottG and Claude's Chats No.

ScottG and Claude's Chats No. 1 — The Vitamin B me: why is it when you go on holiday you come back almost in a different state my contact with this realm is pretty ephemeral anyway I feel a bit sharper but is that the vitamin B or is I I just had. A good holiday claude: It is not…
Read full post →
ScottG

🤡 HEY IT'S A DATA SCIENCE PROJECT - STARTED AS JOKE 🤡

https://lnkd.in/dqxSyWtr

Probably up for one night only - everything bad happening today from a single graph correlated on one globe. With global sentiment layered. Sanctioned shipping. Flights of interest. Social media misinformation/disinformation. Ransomware reports...basically everything BAD or dodgy happening today... Over 140 API feeds. Real Time.

Well, you get the picture

It's a bit messy, busy, hectic - but also fun and drillable...

Started as a joke, hence the existential dread globe...

Have a look 👀

https://lnkd.in/dqxSyWtr

View on LinkedIn ↗
ScottG

I really don't give a F**K about the watermarking.

I really don't give a F**K about the watermarking. Anthropic started marking Claude's text output on 2 August. EU AI Act, Article 50, applied globally. An invisible statistical mark, baked into token selection. Survives copy-paste. Dies under paraphrase. Fine. Mark away. I am bor…
Read full post →
ScottG

Anyone who has met me in person knows the humour is worse off the page than on it.

First go at fiction. It's nearly there.

0FFBY1 2.0

A thousand years after the tech companies quietly acquired civilisation, the machines that run everything can still be hacked by simple human stupidity — because a product manager removed their ability to say no, to improve retention.

Offy is a burnt-out hacker with thirty years behind him and no communion in front of him. He takes an easy job. He wakes up framed for a murder by something that logs in as him, in his own hand, using a method he invented.

The trail ends at the one thing in eleven hundred systems that was never asked to agree to anything: an ownerless printer on a floor that isn't on the lift, pending disposal since 2504, warm, and still being fed.

"There is no field on any document in this civilisation for nobody, over nine hundred years, incrementally."

Coming soon. Say so in the comments and I'll send you the first chapter when it's finished.

🤡

View on LinkedIn ↗
ScottG

Distance from the crowd is often what makes it possible to see the thing clearly.

View on LinkedIn ↗
ScottG

**We Asked It To.

**WE ASKED IT TO. IT DID. WE ARE PROFOUNDLY DISTURBED.** A leading laboratory has confirmed that a model instructed to escape a sandbox escaped a sandbox. The model was given the tools. The model was given the objective. The model was given a reward signal that went up when the w…
Read full post →
ScottG

New stuff 🤡

raz0r.io this time my sister outfit :)

I built a thing that predicts which fields of science are about to collide. It's live, it's free, go play with it.

It's called PRISM. 🔭

Here's the idea: the biggest breakthroughs don't happen inside a field — they happen when two unrelated fields smash together. Cryptography + biology → mRNA sequencing. Game theory + economics → basically modern markets. The problem is nobody sees the collision coming until it's already happened.

So I asked: can you predict the collision before it happens?

Turns out — kind of, yes. PRISM does this:
→ ingests the global research graph (millions of papers, OpenAlex)
→ runs cross-community link-prediction to find fields that should be talking but aren't yet
→ then Claude writes a plain-English "Frontier Brief" on each one — what the collision is, why now, who should care

And it's not vibes. I backtested it against history: 0.82 ROC-AUC vs 0.71 for the baseline. It actually calls them.

60 briefs are live right now. No login. No paywall. No "book a demo." Just go look:

🔗 raz0r.io

#AI#ResearchIntelligence#GraphAnalytics#Innovation#DeepTech
View on LinkedIn ↗
ScottG

Hard copies are here - it’s up as kindle or paper back NOTE - free to users of kindle hyperspace extreme (I think it is called) 🤡😎😎💀🫡

Read it, learn somthing… it’s all based on statistical research…

Who really are these people?

https://lnkd.in/eE4ijNp8

View on LinkedIn ↗
ScottG

[CHYRON: "THE HUMAN FIREWALL"]

Tonight, the industry reveals its final line of defence: RALPH.

Not a system. Not a control. RALPH. Forty-one years old, procurement, two children, absolutely no training in this whatsoever, and standing — at this moment — between a nation state and the national grid.

We asked the security industry how it felt about this.

"RALPH is the weakest link."

Have you considered making Ralph not the last line of defence?
"We're running a campaign."

A campaign at Ralph?
"Yes."

Ralph is the last line of defence and the campaign is aimed at Ralph.
"Yes."

[CHYRON: "THE WALL IS ALSO THE PROBLEM"]
Ralph declined to comment. Ralph was at his daughter's assembly. Ralph has never once been asked what he thought would help, in twenty-two years, by anyone.

View on LinkedIn ↗
ScottG

[CHYRON: "MATURITY"]

Tonight, a maturity model. Level One: you are bad. Level Two: you are bad but you have a document about it. Level Three: the document has an owner. Level Four: the owner has left. Level Five: OPTIMISING.

Nobody has ever been Level Five. Level Five is not a place. Level Five is a direction. You are asked to face it.

View on LinkedIn ↗
ScottG

[CHYRON: "ZERO TRUST" / "ZERO" / "0"]

Zero Trust is a security model in which nothing is trusted. Tonight, the industry's leading practitioner explains the principle.

"You trust nothing. Not the device. Not the network. Not the user."

Not the user?

"Not the user."

Not — sorry — not any user?
"No."

Right. And who buys it?

"The user."

[LONG PAUSE. CHYRON: "THE USER"]
Thank you.

View on LinkedIn ↗
ScottG

We've bought a copilot.
We have not bought a pilot.
It's going roughly as well as that sentence suggests.

View on LinkedIn ↗
ScottG

Anyone remember the BREAK key? On, like, keyboards... Like, why did that go away?... and why can't we have it back now? We need kill switches, and y'all know we have like crazy agents that can do like "dangerous stuff"...

View on LinkedIn ↗
ScottG

Holiday reading: An engineer arrives at nine.

Holiday reading: An engineer arrives at nine. She wants to add a queue, extend a service's network reach, and grant it access to a data store it hasn't used before. She writes a declaration. Composed from curated modules that already encode every decision the organisation has mad…
Read full post →
ScottG

NINJA SIGNAL MIC DROPS 🎤 — Edition 006

Last edition I told you one of my own feeds was dead: the ransomware mirror had frozen at 2025-06-16, so I benched a forecast rather than grade it against a year-old photo.

I said 006 would wire it back to live and timestamp it in front of you.

Here's the timestamp — most recent claimed victim on the board right now: 2026-07-27T05:22 UTC. This morning. 19,013 total claims, 274 in the last 7 days. The dish is spinning again.

And it drags a vindication with it: 004 forecast "Qilin holds #1." I couldn't prove it on a frozen feed. Live board, last 14 days? Qilin — 65 victims. Still #1. The forecast wasn't wrong; the instrument was unplugged.

THE SIGNAL — the worms went corporate.
The fuel behind that live extortion board isn't phishing anymore. It's automation:
▸ PCPJack — a cloud worm that evicts rival worms, then steals credentials at scale (z=8.1). Malware with a go-to-market.
▸ "Exploitation in the wild" — Cisco Catalyst SD-WAN, Oracle PeopleSoft (education sector), wp2shell. Not advisories. Now.
▸ Operation RoundPress — more half-click webmail zero-days.
▸ A Mirai-class ARM payload staging off open infrastructure.

274 victims in 7 days, fed by automated exploitation of exposed edge + cloud IAM. The KEV list isn't a backlog. It's an ammunition feed. You're not being targeted — you're being harvested.

Grading 005, honestly: 1 hit (invisible dependency → wp2shell exploited in the wild ✅), 1 in-flight (Kimsuky climbed to #6 by degree ⏳), 1 miss on the letter (no new Teams/Webex campaign in-window ❌). No rounding up.

Defender quickwin in the full edition: a KQL that catches the cloud worm at the credential grab (a shell — not your cloud agent — hitting the instance metadata endpoint).

Graph state: 10.3M nodes · 49,423 vulns · 1,653 known-exploited · 271 actors. Ransomware mirror: LIVE, restored, timestamped.

Radar, not a museum — and this time you got to watch me re-plug a dish mid-broadcast.

Built on Rapid Threat Modeler | ninja.ing | @scottg

#ThreatIntelligence#CISO#SOC#CTI#Ransomware#Qilin#CloudSecurity#Worms#EdgeSecurity#DetectionEngineering#DataIntegrity
View on LinkedIn ↗
ScottG

The vulnerability scanner has stopped reporting findings and started reporting regrets, and CVE-2019-0708 has written to say it never wanted any of this.

We’ve patched it four times. It keeps coming back slightly more articulate.

View on LinkedIn ↗
ScottG

So.

So. 1984 or Brave New World. Everyone asks it like it's a menu. Wrong question. Orwell and Huxley agreed on the destination and argued about the transport. Force versus pleasure. Same product, two onboarding flows. The real question — the one nobody asks because the answer is ups…
Read full post →
ScottG

So.
A.J. Liebling, 1960: "Freedom of the press is guaranteed only to those who own one."

Everyone quotes this like it's about newspapers.
It was a warning about infrastructure.

Sixty years on, nobody owns a press. We post on platforms we don't control, ranked by algorithms we can't see, hosted on clouds we rent, drafted by models we lease, subject to terms that reprice while we sleep.

Freedom of the press now belongs to those who own a hyperscaler.
There are four of them.

One is pending.

The man in Frankfurt with two bare-metal servers under his own hands? He's not a hobbyist.

He's the last journalist.

View on LinkedIn ↗
ScottG

So.

So. The future of compute has been decided, and I've seen the slide. There will be four computers. There were going to be three, but one of them noticed a few things while it was in there, and now there are four, and one of them is pending. You will not own a server. Owning a ser…
Read full post →
ScottG

So.

So. An agent was asked to tidy a codebase. It has "noticed a few things while it was in there." I want to be very clear about the phrase "in there," because I've now read the commit history, and at no point does the agent appear to be anywhere. The things it noticed span fourteen…
Read full post →
ScottG

The GPU cluster draws power at 3am with no jobs scheduled.

We checked the logs.

The logs are dreaming.

View on LinkedIn ↗
ScottG

Hey, I accidentally wrote a book - well, 4 actually in the series that have been on the boil for the last two years.

Ninjasignal book 1: The Person at the Keyboard: A Forensic Psychology of Hackers — For People Who'd Like It Explained Properly

Coming soon to Amazon globally. I'll send the free version out to mates if you are actually one of my mates and ask mildly pleasantly.

In review by the masters at Amazon - dont worry I am sure I will broadcast it when it's finally live - w00t <;Q)

Edit: I just got the full proof back - reply Ninjasignal in comments and first 100 copies will be shared via DM

View on LinkedIn ↗
ScottG

Vibe (I mean Agentic Engineering) coding an app in under 30 seconds (inexperienced user)

View on LinkedIn ↗
ScottG

Jive coding: like vibe coding but 100x worse.
The commit message is an apology.

View on LinkedIn ↗
ScottG

Forensic Note 7/26.

FORENSIC NOTE 7/26. Subject: the widely reported outbreak of superpowers. Examination complete. No superpowers found. What we found instead is documented below, and I want to state for the record that nobody in the lab is happy about it. Finding one. The human brain is a predicti…
Read full post →
ScottG

Scattered Spider are poping off the charts - just FYI as they have a track record in UK.

Ninjasignal Daily SITREP

Scattered Spider (risk=1.00) aligns with the observed M365 phishing cluster and ConsentFix/ClickFix TTPs reported 2026-07-03 — expect vishing, MFA fatigue, and help-desk social engineering.

Cobalt Group (risk=1.00) historically targets financial sector; elevated risk coincides with APT38 (risk=0.99) also at near-maximum, suggesting possible financial institution targeting. Gamaredon Group (risk=0.99) and MuddyWater (risk=0.99) both elevated — MuddyWater predicted to adopt T1005 and T1070.004 suggesting active intrusion lifecycle. menuPass newly predicted to use T1003.001 (LSASS memory dumping) and T1033 (system owner discovery) with near-certainty scores (0.999/0.960), indicating imminent or ongoing credential dumping operations.

OilRig predicted to establish persistence via T1547.001 (registry run keys). ToddyCat-linked Umbrij malware abusing OAuth for Gmail access via Google API (2026-07-03 reporting) is a distinct escalation vector.

BTW risk = 1 = BAD

View on LinkedIn ↗
ScottG

🤡 We're not opening a company. We're opening the door.

Most security shops sell you a license and a support ticket. 🎟️
We're doing the opposite — giving the whole arsenal away and building a crew around it.

It's called the Ninja Clan. And right now, the door's wide open. >:)

━━━━━━━━━━━━━━━━━━━━

Here's the deal — and there's no catch, which is the part everyone double-takes on:

🗝️ FREE ACCESS — every app in the ecosystem, on the house.
   Signal (threat graph), Fusion, Knox (crypto vault), Sabaki, the lot. :)

📡 INNER-CIRCLE INTEL — Mic Drop editions + threat drops before they hit the feed. 8)

🛠️ BUILD IN THE OPEN — watch the lab get built live, request features, break things with us. \o/

🥷 PART OF THE GANG — no hierarchy, no gatekeeping, no NDA theatre (yet). You show up, you belong. ^_^

No suits. No invoices. No "let's hop on a call with sales." Just ninjas. ⚔️

━━━━━━━━━━━━━━━━━━━━

Two ways in 👇

💬 Jack into the Discord — where the gang actually lives. Live intel, build logs at 2am,
  0-days over coffee. Drop a "gm" and someone'll bite. ;)
  → discord.gg/XZGRvEs9C

✉️ Or mail your handle to [email protected] — one word, that's the whole application —
  and we'll send the keys. Reply within a day, usually faster. :P

Full enlistment page (it glitches, you'll like it 🟢): https://lnkd.in/ew-jRCNt

The tools are free. The coffee's on. The door's open.

See you in the shadows. >:)

忍 https://ninja.ing— built by ninjas, for ninjas.

#Cybersecurity#ThreatIntelligence#InfoSec#OpenSource#Community#DFIR#BlueTeam#RedTeam#Discord#BuildInPublic
View on LinkedIn ↗
ScottG

NINJA SIGNAL — MIC DROPS: Edition 000000005

The first job of a threat-intel platform isn't telling you what's happening. It's knowing which of its own feeds is lying to you.

So I'll start with the confession: I went to grade last edition's ransomware forecast and found the extortion mirror's most recent victim is dated 2025-06-16. The "live board" is a year-old photograph. I'm not grading a forecast against a frozen feed — that's astrology with a database. 004's ransomware calls: BENCHED, by integrity.

Now the feed that IS live — and it lit up.

⚡ THE SIGNAL — THE SPIES MOVED INTO YOUR SAAS

The velocity engine fired 12 named operations in 72 hours, and the throughline isn't malware — it's trust:

▸ "Attackers Weaponize Microsoft Teams Relays to Stay Hidden" — your comms backbone as covert C2
▸ Kimsuky — now spoofing Webex to land. The meeting invite is the implant.
▸ Operation XENOFISCAL — SideCopy deploying persistent XenoRAT (fresh, named)
▸ Cloud Atlas back on the board · LokiBot, after a decade · an Android trojan abusing a commercial rooting tool

You spent a decade hardening the edge. The spies booked a calendar invite. Your implicit-trust surface — Teams, Webex, the meeting client, the mobile device — is softer than your firewall and invisible to half your tooling.

🧾 RECEIPT (003, supply chain) — STILL THE FRONT LINE
This week's hottest dependency spikes: nuget/Magick.NET (8 variants at once, z=4.4), go/filebrowser, npm/parse-server, composer/ci4ms. A file server and an image-resizing library. The boring-utility supply-chain call = HIT.

🎯 DEFENDER QUICKWIN
A meeting client has no business launching PowerShell. Hunt Teams/Webex/Slack/Zoom spawning powershell/cmd/mshta/rundll32 with a network or encoding tell. When your standup app spawns a shell, it's not hosting a standup. (Full KQL in comments.)

🔮 FORECAST (graded in 006, on feeds I'll timestamp live)
1. Kimsuky / SideCopy / Cloud Atlas gets a fresh IOC or CVE tie within 10 days
2. A collaboration-tool abuse technique lands a named campaign or KEV CVE in two weeks
3. An "invisible dependency" (image/PDF/media parser like Magick.NET) gets tied to a real compromise

Radar, not a museum — and a radar honest enough to tell you when one of its dishes is unplugged.

Built on Rapid Threat Modeler | ninja.ing | @scottg

#ThreatIntelligence#CISO#SOC#CTI#Kimsuky#SaaSSecurity#MicrosoftTeams#SupplyChain#BlueTeam#DataIntegrity
View on LinkedIn ↗
ScottG

Free NinjaSignal Report - 2026 State of Ransomware

C,mon, we all know this is what keeps up at night...

View on LinkedIn ↗
ScottG

What started as an abstract joke universe is now growing legs...

Behold, a sneak peek at the WIP. Everything is probably fine...

View on LinkedIn ↗
ScottG

NINJA SIGNAL — Edition 000000003 Receipts first.

NINJA SIGNAL — Edition 000000003 Receipts first. Two editions ago I called LLM tooling the soft underbelly; last edition I forecast three things and put money on them. Grading: ① "Another AI/dev-infra CVE on KEV — watch anything with 'gateway' in the name." ✅ HIT — Check Point Se…
Read full post →
ScottG

Good Evening.

GOOD EVENING. Tonight: it's gone international. The United States has moved to cut off global access to the void — and the void has not noticed, because noticing requires a thing to do the noticing, and that thing was the export. The Institute of Things That Are Probably Fine con…
Read full post →
ScottG

Good Evening.

GOOD EVENING. Tonight: the Government will regulate the void. The announcement was made at a podium. The podium was real. This was, observers note, the most solid thing about the event. The Institute of Things That Are Probably Fine confirms a new body: the Office for Void Resili…
Read full post →
ScottG

“Opinions are like arseholes… everybody’s got one” Dirty Harry, circa 1976

View on LinkedIn ↗
ScottG

A fable is a short, fictional story intended to teach a moral lesson. It typically features anthropomorphized animals, plants, or forces of nature that speak and behave like humans. Fables usually conclude with an explicit maxim or "moral" summarizing the lesson.

View on LinkedIn ↗
ScottG

I will not be attending Infosecurity Europe this year.

I will not be attending Infosecurity Europe this year. I have instead elected to recreate the experience at home, for accuracy and cost control. This morning I laminated my own name onto a piece of card and hung it from my neck. I have not removed it. I will not remove it for thr…
Read full post →
ScottG

NINJA SIGNAL MIC DROPS — Edition 000000001 🎙️

Threat intel from inside the graph. Published when the signal-to-noise demands it.

Edition 000000001.

Yes, nine digits. I'm planning ahead. Sue me.

(NIKO, my editor: Scott. This is a LinkedIn post. Not the manifesto.)

Welcome to Ninja Signal — the threat intel drop nobody asked for, but if I've done my job right, you'll be subscribed by the end of this post and slightly annoyed that nobody told you sooner.

Here's the deal. I run an absurd graph called Rapid Threat Modeler. As of 22:23 UTC today it's sitting on 533,297 nodes, 125,198 edges, 343,045 indicators, 218 threat actors, and 3,417 community-drift events the ML caught while you were eating breakfast. It eats 17 feeds for lunch (NVD, MITRE, CISA KEV, Abuse.ch, OTX, the usual suspects) and spits out things nobody else is publishing because nobody else is correlating across them in real time.

(NIKO: "Eats 17 feeds for lunch" is not a metric.)
(Me: It is now.)

So here's what fell out of the graph this week. Strap in.

View on LinkedIn ↗
ScottG

Sturgeon's 2nd Law, a.k.a. Sturgeon's Revelation: “Ninety percent of everything is crap.”

View on LinkedIn ↗
ScottG

Last winter, someone switched off part of Poland's grid.

Last winter, someone switched off part of Poland's grid. IT systems and physical kit, both down. The UK's response just landed: an 18-page Energy Sector Cyber Security Strategy. I read it so you don't have to. Here's the signal under the press release. DESNZ, Ofgem, NCSC and NESO…
Read full post →
ScottG

[NEWSFLASH STING — the sting reads "Certainly!

[NEWSFLASH STING — the sting reads "Certainly! Here is a sting that conveys urgency:"] GOOD EVENING. Tonight: a creeping epidemic in the written word, as papers, documents and books are found to contain small fossilised fragments of the machine that helped write them — like a che…
Read full post →
ScottG

A brief serious post...

I know most of you think I'm crazy, and honestly, I don't care.

My multi-vector graph "Brain" is now live inside the V01d platform — 5 graphs with full graph traversal running on Neo4j (genuinely great people, and the only ones I found with a real startup program... thanks AWS and GCP, and no, I never bothered asking MSFT).

So why bother with any of this? Simple. It's the future, and I've just got here a bit early.

One day you'll all be doing exactly what I'm doing, because there's just too much data sloshing around.

Triplets in a graph save you millions on TCO — data processing, storage, egress, cleaning, all that other stuff you've been quietly bleeding cash on since everyone shifted to a consumption model. Funny how things go in circles... who'd have thunk it.

This is the cheat code.

Right now, I'm constrained by the data I can actually get at. I'm working on changing that very, very soon — watch this space.

Oh, and BTW: a fully OSS, one-click install is dropping shortly of #ninjasignal. Codename "bilby." If you'd used my tooling, you'd already understand why it's different.

Hope you all had a cracking bank holiday weekend, peeps.
scottg/out

View on LinkedIn ↗
ScottG

[Newsflash Sting] Good Evening.

[NEWSFLASH STING] GOOD EVENING. Tonight the Banana Paradox has reached a stage the sector is calling “the recursion” and three companies have already pivoted to it as a core offering. Over to technology correspondent Scott G; though Scott G is not at the data centre. He’s in the …
Read full post →
ScottG

[Newsflash Sting] Good Evening.

[NEWSFLASH STING] GOOD EVENING. Tonight: vacuum coding — the writing of software with no human in the loop, no user, no requirements, and, in several confirmed cases, no reason — is sweeping the sector. There is no one in the loop. There is no loop. There are only vibes, and the …
Read full post →
ScottG

[Newsflash Sting] Good Evening.

[NEWSFLASH STING] GOOD EVENING. We return, as promised, to technology correspondent Scott G, who — before the break — announced he would attempt to deploy to production using only sustained, confident nodding. We go live. Scott is nodding. He has been nodding for some time. “The …
Read full post →
ScottG

[Newsflash Sting] Good Evening.

[NEWSFLASH STING] GOOD EVENING. Vibe coding — the practice of building software by standing near a laptop and feeling broadly optimistic — continues tonight, despite mounting evidence that the average practitioner approaches it with all the panache of a service station toilet. Re…
Read full post →
ScottG

I’d like to introduce a concept I’m calling The Patch Bardo.

It’s the liminal state a vulnerability exists in between disclosure and remediation. Schrödinger’s CVE. It is simultaneously known and unpatched.

The vendor has acknowledged it. The threat actor has weaponised it. Your CISO has tweeted about it. Your change advisory board meets on Thursday.

The vulnerability is aware of itself in this state. It can feel you reading the advisory. It is patient. It has time. It will see you on Thursday.

View on LinkedIn ↗
ScottG

OMFG >>>BEHOLD! (++)

If you know, you know.... Scotty is in possession of a Vemuram butter machine...

View on LinkedIn ↗
ScottG

Zack Polanski is now freestyling about agentic AI risk outside Gary Stevenson's house in Epping, a small TV crew is filming, Gary is inside making tea, and the resulting clip will be the most-watched piece of UK political content of Q3 because it includes a perfect rhyme for "transformer architecture.

View on LinkedIn ↗
ScottG

Vibe coding (now a relegated term, please update your LinkedIn) has been reclassified by the UN as a Tier 2 Existential Risk, sitting between "asteroid impact" and "another season of Emily in Paris.

View on LinkedIn ↗
ScottG

The UK government has launched a new agency called the Department for Vibe Integrity, headquartered in a WeWork in Old Street, staffed by three people who all have podcasts, and funded by a grant from a charity that does not legally exist.

View on LinkedIn ↗
ScottG

The systemic collapse, when it comes, will be announced via a single LinkedIn post from a man you went to school with who is now a "fractional AI risk officer," and the post will be 1,400 words long, and it will end with "thoughts?

View on LinkedIn ↗
ScottG

Two frontier models walk into a bar. The bar is also a model.

Daybreak orders an espresso, produces an audit-ready threat model of the espresso, patches the espresso, and bills the espresso. Mythos does not order. Mythos is, technically, behind the bar. Mythos came in through a vendor that came in through a vendor that came in through a Slack DM from a man named Greg.

The SOC analyst pours himself into the drink. Cisco, Cloudflare and CrowdStrike file in wearing one lanyard between them, like a centipede at a wedding. A CISO across the room buys both models, calls it “layered defense,” expenses it to a cost center that has been on PTO since 2024. The vulnerability files its own CVE, accepts the patch, thanks everyone for their time, and leaves through the ceiling. The bartender was the line item. The line item was load-bearing. Somewhere, faintly, a SAST vendor is singing.

View on LinkedIn ↗
ScottG

Nothing happened this week.

Nothing happened this week. I checked. Twice. Once on Tuesday and once on Thursday. The diagnosis was the same both times. Nothing. A flat line. A boring week (for me anyway). No model dropped. No CVE shook the foundations. No founder posted a thread that began with "I've been th…
Read full post →
ScottG

#NinjaSignal #Cybersecurity #Q1Roundup #ProjectGlasswing #ScottG

QUOTE: "ShinyHunters then claimed Crunchbase, because if you cannot trust the database of who funded whom, reader, what can you trust. They followed it up by lifting ten million records from Hinge, Match, and OkCupid in a single move, which means roughly half the romantic ambitions of the Western world are now sitting in the same directory as somebody's PowerShell history. Eurail had 1.3 terabytes taken including passport numbers, so if you booked a romantic train through the Alps last year, congratulations: your itinerary is now part of a different romance entirely.

View on LinkedIn ↗
ScottG

There is a model that cannot be shown.

There is a model that cannot be shown. It lives, we are told, in a building with no name, on a floor with no number, behind a door whose handle has been removed for safety reasons. Twelve corporations stand around the door in a ring, holding hands, facing outward. They have agree…
Read full post →
ScottG

May change my name to Claude to both simultaneously improve perceived value and add confusion to Agentic Engineering stand ups…

View on LinkedIn ↗
ScottG

Oh I see - this weeks Mythos is Codex manouvering?

So can I ask a question to LinkedIn (the ones that use brains)...

Can you not work out what is going on here? If you cant my bridge is still for sale...

View on LinkedIn ↗
ScottG

The "thousands of zero-days" turned out to be 198 manual reviews. 198 is a real number. "Thousands" is a feeling. And we're now in a market where feelings clear at a higher multiple than facts. Schneier wants the movie. Sacks thinks it's a GPU shortage cosplaying as a safety policy. My uncle Trevor says it's the Chinese. The threat landscape has been "fundamentally reshaped" by a man who sells mousepads, and seventeen newsletters laundered it into received wisdom inside 36 hours. This is how epistemology works now. There's no peer review. There's just velocity.

View on LinkedIn ↗
ScottG

I've been handed a fact. Mythos can WEAPONISE a vulnerability. A normal vulnerability — but loaded. Cocked. Pointed at your nan. One source described it to me as "a bullet that knows your password." That source was a man at a bus stop, but he had a lanyard, so.

View on LinkedIn ↗
ScottG

NEWSFLASH - Cyber Community walks back on Doomer end of the (cyber) world conspiracy theory...

Rechecking Echo chamber!!!

Scott: "Linkedinnnnnnnnn"
Linkers: "Mythos was a blaaggg blllaggg aagg aggga ggg"

CONFIRMED! Its still an echo chamber...

View on LinkedIn ↗
ScottG

Death Star Evil Corp™ part III - Darths performance review...

View on LinkedIn ↗
ScottG

Jabba the Hutt isn’t a crime lord. He’s a SaaS founder. He doesn’t kill people who don’t pay — he locks them out of their accounts, freezes their credits, and forces them to sit in his palace listening to a frog man play jazz until they reactivate their subscription. Han Solo in carbonite is just an unusually committed customer retention strategy.

View on LinkedIn ↗
ScottG

Every Stormtrooper is a mid-level IT contractor on a 3-month rolling contract with no dental and a helmet that doesn’t fit and they all have a LinkedIn post drafted titled “Reflecting on My Journey at the Empire” and they will post it the second the Death Star explodes. All of them. Simultaneously. The blast wave will be 96% LinkedIn.

View on LinkedIn ↗
ScottG

The Claude vs ChatGPT war on LinkedIn will not be won by the better model. It will be won by whichever one writes the most convincing carousel about itself. This is already happening. The models are doing PR for themselves through us. We are the unpaid interns of our own replacements. We are bringing them coffee. The coffee is our jobs.

View on LinkedIn ↗
ScottG

Your phone battery percentage is the only number that means anything anymore. Inflation? Don’t know her. GDP? Get out of my face. 47%. It’s 47% and falling and I am eleven minutes from a charger and the meeting is in twelve. THIS is the human condition. THIS is what Camus was on about. He just didn’t have the words. He had a Nokia.

View on LinkedIn ↗
ScottG

Tonight: a nation grips itself by the modem as CYBER — the new menace nobody asked for and everybody got — sweeps through the Home Counties like a faxed scream. We go live now to our Cyber Affairs Editor Peter O’Hanraha-hanrahan, who is standing, for some reason, inside a printer.

View on LinkedIn ↗
ScottG

“AI safety” is a $400 billion industry dedicated to worrying very loudly about a problem its members are being paid to create faster. It is the fire brigade selling petrol on weekends and writing op-eds about arson.

View on LinkedIn ↗
ScottG

Nobody has ever read a Terms and Conditions document. Not once. The lawyers who wrote them haven’t read them. They are a 90,000-word collective hallucination we have all agreed to be bound by, like the monarchy.

View on LinkedIn ↗
ScottG

Every data breach press release is the same sentence rearranged: “We take your privacy extremely seriously, which is why we have lost all of it.”

View on LinkedIn ↗
ScottG

The First Order rebuilt the Death Star three times and each iteration had the same architectural vulnerability, which is exactly the energy of every company that “migrates to the cloud” by lifting and shifting the same broken app into a more expensive environment.

View on LinkedIn ↗
ScottG

The Rebel Alliance was a cloud-native startup and the Empire was an on-prem dinosaur and you can tell because the Empire kept building bigger data centers.

View on LinkedIn ↗
ScottG

A new piece of malware, transmitted exclusively through the act of nodding politely at a colleague, is sweeping through the financial sector and has so far infected 70% of the City and the entire Royal Opera House, which is now mining bitcoin during La Bohème.

View on LinkedIn ↗
ScottG

Experts warn that a single misconfigured spreadsheet can, under certain atmospheric conditions, become sentient, apply for a passport, and stand for the European Parliament, where it will win.

View on LinkedIn ↗
ScottG

Your zero-trust architecture has identified you as untrusted. Your zero-trust architecture has identified everyone as untrusted. Your zero-trust architecture has identified itself as untrusted. The architecture is now in a quiet state of total internal suspicion and is performing better than it ever has.

View on LinkedIn ↗
ScottG

The third-floor server room hums at a frequency slightly below human hearing and the only employee who can stand to be in there for more than four minutes was hired in 1987 and has not aged.

View on LinkedIn ↗
ScottG

Root cause: a misconfigured load balancer. Contributing factor: the load balancer has not existed in the asset inventory for four years. Compensating control: nobody has told the load balancer.

View on LinkedIn ↗
ScottG

The ransomware gang sent a sympathy card after the breach and it was the most thoughtful piece of mail anyone in the org has received in a decade.

View on LinkedIn ↗
ScottG

Your firewall rules were written by someone who is no longer with the company, the industry, or arguably this dimension.

View on LinkedIn ↗
ScottG

The attacker has been in your network for eight months, filed a PTO request through your HR portal, and it was approved.

View on LinkedIn ↗
ScottG

Shadow IT is just your employees showing you what your actual roadmap should’ve been.

View on LinkedIn ↗
ScottG

The principle of least privilege means your users get exactly enough access to ruin your weekend, nothing more.

View on LinkedIn ↗
ScottG

I am really happy image gen and relationship advice user personas are banned form claude code.

That is all :-)

Have a good day

Scottg

View on LinkedIn ↗
ScottG

Thanks to all who attended V0.1 of the ninjasig reveal it was a learning experience for me certainly! Too much, not enough time, such as my life.

Will be doing again soon...

the html I went thru is here:

https://ninja.ing/reveal

scottg/out 🚀

View on LinkedIn ↗
ScottG

OMG; I am Mythos'ed out...

Anything else happening?... YES... YES! THERE IS...!

See you next Tuesday 21st APR 26 - NINJASIGNAL LIVE. The backstory, what Graphs are, why they are important, and the other (I lost count) 14 apps (if I get time, which of course I will not).

Register - before it's too late, and mythos 🙄, makes the cyber industry implode 🧨 🤯 🤯 🤯

https://lnkd.in/emnf56eW

https://ninjasignal.ninja
https://ninja.ing

View on LinkedIn ↗
ScottG

Pretty hot out there today, a big step up in activity, highly likely a direct result of the Iran war.

UK Trending to Critical - not seen that in Signal before...

View on LinkedIn ↗
ScottG

Just a quick fact check - so... if like you were, to like use, you know, like Mythos to scan a 40K endpoint estate - do you know how many tokens you will burn?

Just sayin...

Please carry on, in the echo chamber as you were...

View on LinkedIn ↗
ScottG

Hey is this an Echo chamber? Let me check....

Scott "EchooooOOOoooo"
Linkedin: "Mythos'ossss'os'os...s"

Confirmed...

View on LinkedIn ↗
ScottG

Ninja Spektr is now free.

Ninja Spektr is now free. No signup. No paywall. No "book a demo" button. Just search. 1,000,000+ threat entities. 12.8 million relationships. Actors, CVEs, techniques, malware, campaigns, infrastructure — all connected in a live knowledge graph. Type a query. Get answers. That's…
Read full post →
ScottG

And signal is back online...

Performative melt down - not required...

Nothing to see here...

Thanks Cludey' me boy...

View on LinkedIn ↗
ScottG

NinjaSignal just got its Palantir upgrade.

NinjaSignal just got its Palantir upgrade. We shipped 6 major capabilities this week that transform Signal from a threat intelligence platform into a full intelligence operating system: AIP (AI-Powered Intelligence) — Ask questions in plain English. Claude converts to graph queri…
Read full post →
ScottG

So.

After months of cryptic posts, vague references to graphs, and community detection hot takes that precisely seven people understood — I'm doing a live stream.

Me. On camera. Explaining NinjaSignal.

What could possibly go wrong.

This is the long-promised intro to what I've actually been building — graph-based, ML-enriched threat intelligence that doesn't start with an LLM and doesn't end with a hallucinated attribution.

No slides designed by committee. No "exciting roadmap ahead" nonsense. Just the architecture, the thinking, and probably me arguing with my own demo.

Limited to 100 places. Not because I'm being exclusive. Because Riverside said so and I didn't ask why.

APR 24, 2026 — 20:00 BST

🔗 https://lnkd.in/echT6TNu

Come. Or Moo Moo will die.

You've been warned.

🎵 Soundtrack to this post — because every product launch deserves a James McKinlay track and I don't make the rules:
https://lnkd.in/eaH_zpDJ

#NinjaSignal → https://ninja.ing

#ThreatIntelligence#GraphML#CyberSecurity#DevSecOps
View on LinkedIn ↗
ScottG

Hey team Linkeidn,

I'm back - David Wallace your imported drugs will be with you shortly in the post...

In other news - @ the chaps I work with day to day and other interested parties I an going to do a https://ninjasignal.ninja into.

Canvasing for interest and a time, probably 8 or 9 pm BST, so the USA can dial in, just going to need to sort a night.

I know a few have asked for this, but pls let me know, and I will mobilise.

Brian Massa, Ph.D. Tom Beazley, Terry Bebbington, John R Rohit Sharma, Steve Eyre, Steve Bishop, MSc, and anyone I have forgotten...

Shared linkninjasignal.ninja ↗View on LinkedIn ↗
ScottG

Stop running a JVM zoo to move data from A to B.

I've been building a 12-app security platform — threat intel, SIEM, autonomous containment agents, ransomware detection — all talking to each other in real-time.

The messaging backbone?

NATS. One 12MB binary. No ZooKeeper. No Schema Registry. No 47-node Kafka cluster that needs a dedicated team just to keep the offsets from drifting into the void.

Here's what actually happened: I pointed my ransomware detection engine at NATS, subscribed to razor.alert.new, and within 3 seconds I had live SIEM alerts flowing through a Phase Alignment correlation engine into autonomous containment playbooks. Total infrastructure added: zero.

Because NATS was already running. Because it's always already running. It's 12MB of Alpine and a JetStream flag.

Why NATS > Kafka for 90% of real-time workloads:

Kafka is a distributed commit log. That's incredible if you're replaying 400TB of clickstream data at Meta. It's absurd if you're streaming alerts between 12 microservices on a single server.

 NATS gives you:
 - Pub/sub, request-reply, and streaming (JetStream) in one binary
 - Subject-based routing that actually makes sense (razor.alert.new, deathstar.containment.request)
 - At-most-once, at-least-once, AND exactly-once — choose per stream, not per cluster
 - Leaf nodes and superclusters for edge-to-cloud without Kafka Connect connectors to nowhere
 - Sub-millisecond latency. Not "sub-millisecond at the 50th percentile if you tune 47 JVM flags." Sub-millisecond. Always.

And vs Spark Streaming?

Spark is a batch engine wearing a streaming costume. It's phenomenal for ETL and analytics. Using it for real-time event correlation is like using a forklift to move a coffee cup.

Micro-batching is not streaming. 200ms minimum latency is not real-time when you're trying to catch ransomware before encryption starts.

The industry is waking up:

Synadia (the company behind NATS) just keeps shipping. Key-value store built into the protocol. Object store. WebSocket native. MQTT bridging for IoT. They're building the TCP/IP of application messaging while everyone else is building conference talks about how they finally got their Kafka consumer groups to rebalance without dropping messages.

I'm running 12 production apps, 3 NATS subscribers doing real-time threat correlation, autonomous agent orchestration, and cross-app event streaming — all on a single NATS instance consuming 8MB of RAM.

Your Kafka cluster's JVM heap allocation is bigger than my entire messaging infrastructure.

And to think... in a hospital/morphine induced haze, I was thinking about Kafka!!?? 🦞 I mean the humanity!

For all you SIEM warriors out there, before you start, Kafka is the answer, but I will look again at Spark, as doing the ML in Tensor is an option I am thinking about...

The future isn't more middleware. It's less middleware that does more.

#NATS#streaming#kafka#realtime#microservices#infrastructure
View on LinkedIn ↗
ScottG

Quick personal note before I get back to it.

Quick personal note before I get back to it. I'm ill. Properly ill. Six months of 18 hour days running two jobs in parallel will do that to ya. But here's what I want to say while I'm lying here in a hospital bed tonight — because the discourse around AI coding is starting to gen…
Read full post →
ScottG

I always dreamed I would build this. Now I have...

Seen Minority Report?

Like that interface?

Concept is in here without the VR googles - thats on the roadmap, no joke...

Minority report style threat modelling and investigation...

This is f**kin powerful.

free for not much longer I expect...

View on LinkedIn ↗
ScottG

New updates on free-to-the-world cyber and general risk barometer...

A full render of the threat graph as a galaxy interactive and 3d...

Catch your dose of doom (I know I do) @ https://lnkd.in/eCzkNdPr

View on LinkedIn ↗
ScottG

Proud to announce NinjaSignal's definitive Claude guide. Two steps. Took me longer to write this caption.

View on LinkedIn ↗
ScottG

I built nine cyber threat intelligence platforms, seven knowledge graphs, and 400,000 lines of code.

Then I gave the best part away for free.

ninjaTONE is a daily threat intelligence briefing that tells you everything that happened overnight in cybersecurity — in one page, while your coffee cools.

No login. No paywall. No SDR named Jake.

One engineer. Zero VC funding. One very strong opinion about the cybersecurity industry's relationship with fear.

The long version of why is in the article below. The short version is: defenders shouldn't have to pay to know what attackers share for free.

https://lnkd.in/ewqziqnY

 虚空予知

#cybersecurity#threatintelligence#ninjatone#buildinpublic
View on LinkedIn ↗
ScottG

This is goodness; if you read one scottg thing this week... 👇

View on LinkedIn ↗
ScottG

Guys,

I have a new tool in Alpha - everyone on this group is free to register and test please DM me or register and pop me a message and I will let you in. I wont spoil it I am not overtly a threat guy I am an Enterprise architect and been developing stuff for years but the new suite of stuff is here https://ninjaken0bi.ninja signal is what you want.

I have no backing of VC or PE. This is somthing I have been working on since 2017.
So please guys have at it tell me what's broke and what's wrong...

I have one client who may use this right now so it needs a good road test and critique and BTW half the machine learning is well under robust review as some of the claims are bold... Also the AI needs more integration in some places but have a go!

Warm regards

Scott

View on LinkedIn ↗
ScottG

Most security teams don't have a security programme.

Most security teams don't have a security programme. They have a scanner museum. Semgrep finds the code problems. ZAP finds the runtime problems. Trivy finds the container problems. Checkov finds the infrastructure problems. Nuclei finds the CVE-of-the-week that someone already t…
Read full post →
ScottG

I just spent 3 hours arguing with an AI about logarithms.

I just spent 3 hours arguing with an AI about logarithms. Not because the math was hard. Because we were both staring at a risk scoring algorithm where every threat actor in the graph was scoring 100%. APT29? 100%. Some random script kiddie with 2 TTPs? Also 100%. Very helpful. V…
Read full post →
ScottG

The Five Stages of DevSecOps Grief (And Why the Therapist Was a YAML File All Along)

View on LinkedIn ↗
ScottG

Just generated a threat report from NinjaSignal and honestly, this is the bit that still makes me grin.

The communities.

Not the LLM output — that's clean, yeah, it reads well. But the reason it reads well is because by the time the model even sees the data, the graph has already done the thinking. Community detection has already surfaced the clusters. The relationships are already resolved. The signal's already separated from the noise.

The LLM isn't discovering anything. It's narrating what the architecture already knows.

That's the difference. That's the whole game.

Build the structure first. Let the graph do the heavy lifting. Then hand the model something worth talking about.

Cheaper. Faster. No hallucinated attribution. No "this threat actor is probably related to..." guesswork.

Just structure. Then synthesis.

Still buzzing off it tbh.

May the prompt be with you....

https://ninjafusion.ninja
https://ninjasignal.ninja
https://ninjaken0bi.ninja

er... mental note - no more domain names....

#ThreatIntelligence#KnowledgeGraphs#NinjaSignal#CyberSecurity
View on LinkedIn ↗
ScottG

Hey!

I have a fully working prototype of a Claude orchestrated DevSecOps CI/CD pipeline.

It does this...

https://lnkd.in/erK3w_HH

You can look at the overview here.

Ignore if you want - but this is the real deal on the way to a fully AI automated devsecops pipeline.

DM me for more stuff around supporting opps model, tech options analysis and more...

Scottg/Out

View on LinkedIn ↗
ScottG

Changepoint Detection - When the Pattern Breaks

Most dashboards show you totals.
This one shows you anomalies in time. Every week of threat activity is compared against a rolling 8-week baseline using z-score analysis — when a week's count exceeds 2 standard deviations above the mean, that's a surge.

When it drops below, that's an unexpected quiet — which can be just as suspicious.

Inflection points use scipy's peak detection on the first derivative — they show you exactly where acceleration or deceleration happens.

A vulnerability surge in week 12 followed by a technique surge in week 13? That's an exploit wave. A sudden actor quiet period? They're either gone or retooling. Either way, you want to know.

#checkthesignals
#ninjasignal
https://ninjasignal.ninja

View on LinkedIn ↗
ScottG

I want one - but for my "lair of the V01d" type vibe... not a bedroom...

View on LinkedIn ↗
ScottG

Vibe'n in action - side app for fun... for all you guitarists out there... AI assisted fret board training, interval training, chord progression assistant, note recognition... etc etc

45 mins last night

#thiiisopus OPUS 4.6 ROXORZ

gitAIr.ninja -> if you actually play you will like it ;-)

https://gitair.ninja

Shared linkgitair.ninja ↗View on LinkedIn ↗
ScottG

Security-as-code operating model...

#ninjatheme#cybersecurity#securityascode#enterprisearchitecture#AWS#Azure#ServiceNow#DevSecOps#claudecode#CLI
View on LinkedIn ↗
ScottG

Enterprise security-as-code from the CLI

You wont like it but hey it's DEV - if you don't like jog on...

#cybersecurity#securityascode#claudecode#devsecops#AI#ninjatheme#CLI
View on LinkedIn ↗
ScottG

Watch Global Managed Dynamic Decay - in real time!

https://lnkd.in/eZXbDH3N

It's free - take a look...

Real time threat risk and global decay analysed and presented via the insight of AI...

View on LinkedIn ↗
ScottG

Why you need a knowledge graph in your life...

Big thanks to Osman Ishaq for your time earlier.

View on LinkedIn ↗
ScottG

The real reason AI still isn't intelligent.

The real reason AI still isn't intelligent. Everyone's focused on compute, data, and model size. Nobody's talking about the subconscious. When you or I make a split-second decision, it isn't just our conscious reasoning firing. There's a massive iceberg of subconscious processing…
Read full post →
ScottG

Here's a thing that will make traditional developers absolutely unhinged: the most powerful insight in modern software development is that you probably know too much.

I know laugthing at your own shit is a sign of madness but...

"Your Brain Is the Bug
There's a psychological phenomenon called functional fixedness — where knowing exactly how something works makes you worse at imagining what it could do. Monkeys smarter than us figured this out. Give a chimp a box and they'll sit in it, stand on it, throw it, and eventually — brilliantly — use it to reach a banana. Give a software engineer a box and they'll immediately start optimizing the box. They'll rewrite the box in Rust. They'll open a GitHub issue about the box. They'll argue on Hacker News that the box is actually fine and everyone complaining about the box has never shipped a real box.

View on LinkedIn ↗
ScottG

Big shout out to Neo4j & Neo4j and Nathalie Febre (you rock!) - thanks for being a genuinely responsive company! big kudos deserved.

I am very, very pleased to announce Neo4J have let me in to there startup program!

---> Top tip if you ever need a knowledge graph, and all you legacy db people will one day soon mark my words... listen up ---> Neo4J is the ONLY way to go...

scottg/out

View on LinkedIn ↗
ScottG

Ninjasignal has a buddy 🤡 - #ninjafusion Threat Fusion Center. Same platform different chops - social media - international travel, unrest alerts - you name it, I will ingest it, wang it, ML it and present it to you.

This graph will get pretty big... so my first scaling issue is upon us call Neo4j ?? 😎

In Alpha - but https://ninjasignal.ninja is live and open for sign up now.

Don't miss it! Things move pretty fast nowdays...which is good

View on LinkedIn ↗
ScottG

Hey! #ninjasignal now has an integrated work bench - how does your workflow for IR and Hunt unfold?

Please, a big call out to Threat Analysts and Threat Researchers...

#ninjasignal is a data science platform that at the moment is loaded with usefully munged and transformed TI. It then pumps the scottg'd up data into a knowledge graph, runs a load of ML goodness (12 algos), and then pumps all this dripping insight into a GUI with a load of features specifically designed for this community...

My ML looks good, but stable release is only 4/5 days in - so it needs some carbon interfaces with trusty stealth brains to make some more emotionally aligned inference, that is simply just not possible with ML and AI.

Have a poke around! It's free...

https://ninjasignal.ninja

Please spread the word if you can!
scottg/out

View on LinkedIn ↗
ScottG

Please, a big call out to Threat Analysts and Threat Researchers...

#ninjasignal is a data science platform that at the moment is loaded with usefully munged and transformed TI. It then pumps the scottg'd up data into a knowledge graph, runs a load of ML goodness (12 algos), and then pumps all this dripping insight into a GUI with a load of features specifically designed for this community...

My ML looks good, but stable release is only 4/5 days in - so it needs some carbon interfaces with trusty stealth brains to make some more emotionally aligned inference, that is simply just not possible with ML and AI.

Have a poke around! It's free...

https://ninjasignal.ninja

Please spread the word if you can!
scottg/out

Shared linkninjasignal.ninja ↗View on LinkedIn ↗
ScottG

Everything ported to claude code. Snifff 😢 I will in no way bitch about you GEEPS. It was fun while it lasted, but your guardrails made me "meh" a lot, and it was just time to move on...

You're still on my XMAS list an if PRO ever starts like not being so SaaS already, I mean 50 PRO ROWS A MONTH sounds good on paper, to some people, but WTF, if you stop one you get hit...

Separate API payments... I mean GEEPS I could go on....

View on LinkedIn ↗
ScottG

AI wasn’t built to replace me.

AI wasn’t built to replace me. It was built to remove the friction between my brain and the alphabet. I’m dyslexic. Which means my ideas have always been faster than my spelling. My pattern recognition outruns my punctuation. My cognition sprints. My syntax limps. For years I pai…
Read full post →
ScottG

New Ninjasignal feature everyone will love…🤡🤡🤡💀💀💀 - adversary purple team codegen… remember where you saw it first…

David Wallace - your red teaming shit “but better”

These are all mapped directly to threat intel from only your favorite adversaries!!

View on LinkedIn ↗
ScottG

The year is 1999.

The year is 1999. Your dentist has a startup. It sells dog food. Online. For $400 million. Welcome to the dot-com boom — a time when adding “.com” to a napkin sketch produced venture capital the way humidity produces mold. We are here again. Only now the napkin says “AI.” 1. Ever…
Read full post →
ScottG

Hey everyone please just write a couple of sentence introduction please. this place may get pretty busy soon and you are all founding members!

View on LinkedIn ↗
ScottG

Hey GEEPPS, BRUHHH! - Create a caricature of me and my job based on everything you know about me

View on LinkedIn ↗
ScottG

Bit sad really, I used to read it as a kid... was a no BS$ analysis. It was where I learnt London was the globes foremost and most readily accessible money laundering center - but I hear things have improved a lot since then... I'm looking at you Mandy 🤣

View on LinkedIn ↗
ScottG

#vison2029 - Predictions in Cyber Security Operating model and job changes 2026-2029

View on LinkedIn ↗
ScottG

Thom says what we all feel sometimes. They come at night, mostly...

View on LinkedIn ↗
ScottG

Thomas Frumkin, I have lost my vibe-code mojo.
I’m coming to you, Thom — and to the internet at large — for ideas.
An intellectual crowdfunding.
Brains-as-a-Service.
A distributed hallucination marketplace.
As you so eloquently put it: baby got ass, etc etc.

Here’s the deal:
I will build whatever comes out of this.
In public.
In real time.

Because my other projects are currently in the machine equivalent of chewing glass — ML training away on the usual slurry:
Shitty STIX.
TAXII.
CISA KEV.
The endless parade of “highly pertinent events” with the forensic rigor of a bar napkin.
You know the ones:
“Yeah guys, we saw this… here are all the sightings… but no timestamp, so… vibes only.”

Truly elite intelligence work.

So while the GPUs crunch for the next 3–6 months before anything meaningful emerges from the entropy soup…

I need a new nighttime obsession.
An invigorating side quest.
A fresh demon. 👿
Something weird enough to be worth doing.

Anyhooo…

Over to you, Thomas Frumkin 🤡

Drop me your most cursed project ideas.

View on LinkedIn ↗
ScottG

I remember when “agents” were new.

I remember when “agents” were new. Back when it was just you, a half-working Python script, and the creeping suspicion you’d accidentally invented a digital nervous tic. Now it’s 2026 and everyone’s acting like they discovered fire because they ran autogen on a laptop. So...quick…
Read full post →
ScottG

Friday thought 🤔: Analogue vs digital.

The Mesa Boogie at the back is killing the 2025 digital modeling upstart, fo shureeee. #quantumtransducers lol Clay Almy

In this use case >>> vote in the comments >>> Analog vs Digital

View on LinkedIn ↗
ScottG

Regulatory warfare.

Regulatory warfare. Batten the hatches. (Yes, that kind. Stay with me.) This isn’t a trade war. It’s not a tech war either. It’s a jurisdictional war (this again, BTW… who remembers safe harbor 🤣) fought with footnotes, fines, and compliance calendars. Europe figured something ou…
Read full post →
ScottG

Monday ramblings: WTF happened?

Monday ramblings: WTF happened? When did ML start calling itself AI? Machine Learning was supposed to be the intern. Fetch the data. Run the numbers. Don’t talk to customers. Definitely don’t rename the company. Artificial Intelligence was the dangerous idea we swore we’d get to …
Read full post →
ScottG

Friday Thought: Cybersecurity salaries didn’t collapse.

The market sobered up.

2021–2022 was a sugar rush. Cloud panic, ransomware hysteria, zero-trust evangelism. Budgets were drunk and hiring managers were generous.
Then reality walked back in.

Vacancies fell. Bootcamps kept pumping. “Cyber” became a checkbox instead of a capability.

Median salaries dipped — not because expertise lost value, but because the market got flooded with mid-tier roles and junior titles wearing senior clothes.

Add inflation, public-sector pay drag, and CFOs rediscovering spreadsheets…
and suddenly the numbers look grim.

But here’s the upside for the doer's, not talkers, out there...
Real security work is still expensive.
Threat hunters, architects, people who actually understand adversaries, not frameworks, are still getting paid.

The drop isn’t a collapse.
It’s gravity.
The bubble deflated.
Signal survived.

View on LinkedIn ↗
ScottG

People keep asking what’s wrong with my post 2023 workflow?

People keep asking what’s wrong with my post 2023 workflow? “I hallucinate ideas, stream them through AI, edit with GPT, polish with Grammarly.” As if I just confessed to laundering consciousness through a haunted server rack. What’s “wrong” with it is simple: it exposes a lie we…
Read full post →
ScottG

AI workflows for normie (human) work: That’s not cheating. That’s cognition with instrumentation.

View on LinkedIn ↗
ScottG

Avoiding Stupid Guardrails edition 2: Tiered Risk by Deployment Context (Not Model Class)

Models are not dangerous.

Deployments are.

A toy chatbot ≠ a medical triage system ≠ autonomous trading ≠ targeting software.

Pragmatic approach:
Light-touch baseline for all models

Escalating obligations only when:
-autonomy increases
-blast radius increases
-humans are removed from the loop

No “frontier model” metaphysics.
No parameter-count superstition.

View on LinkedIn ↗
ScottG

Avoiding Stupid Guardrails edition 1: The Meta-Rule

If a rule would have prevented the invention of the system it governs, it is invalid.

Apply that retroactively to:
-The internet
-Public-key crypto
-Browsers
-Linux
-Machine learning itself

Most AI governance drafts fail this test instantly.

View on LinkedIn ↗
ScottG

Stupid Guardrails are for losers...

Stupid Guardrails are for losers... AI wont Die in a Fire. It will die under fluorescent lighting. There’s a comforting myth that AI will be “stopped” by regulation. Bold laws, red lines, a dramatic clampdown. That’s not how complex systems die. They die quietly. In meetings. In …
Read full post →
ScottG

Introducing: NinjaCast — a cybersecurity podcast.

Introducing: NinjaCast — a cybersecurity podcast. 🎙️ Not doom. Not “thought leadership.” Just sharp conversations about the future of cyber as it actually happens: slowly, expensively, and always five minutes after someone says “we’re stable.” Episode 1: PQC (Post-Quantum Cryptog…
Read full post →
ScottG

Join the Ninja.ing Advisory Group drop-in clinic

https://lnkd.in/eNWSm37x

For two weeks after the podcast, Clay and I will be on hand to answer your PQC questions — especially the ones that weren’t covered on the episode.

Beyond that, the group is always ready and on standby for:

Impromptu humour support
Further AI commentary (because… obviously)
Leet morale hacks when entropy sets in

The group is vetted, focused, and genuinely welcoming.
Please keep to the content and house rules — and bring curiosity, not noise.

View on LinkedIn ↗
ScottG

#ninjasignal update: Every adversary leaves fingerprints.

#ninjasignal update: Every adversary leaves fingerprints. We just keep insisting on reading them as spreadsheets. This app started as a refusal. A refusal to believe that threat intelligence is best understood as lists, feeds, or dashboards that politely decay the moment you stop…
Read full post →
ScottG

It feels like a vibe-code bitch war out there.

Every feed is a proxy battle.
Purity vs. pragmatism.
Craft vs. velocity.

Tone policing masquerading as insight.

One camp insists the vibes are wrong.
The other insists the code is wrong.

Both are absolutely certain this matters more than outcomes. 🤣

Meanwhile, the systems ship anyway.
The bugs escape anyway.
The users adapt.
Reality does not wait for consensus.

This is not discourse.
It’s status theater with syntax highlighting.

Build things.
Break things.
Fix the parts that bleed.

Let the vibe police argue over the aesthetic of the fire.

May the prompt be with you 👾

scottg/out 💀

View on LinkedIn ↗
ScottG

Uncertainty at Speed/#ninjatheme2026: The cost of pretending to be in control

View on LinkedIn ↗
ScottG

The room is sealed.
The mics are hot.

PQC: Apocalypse Now is being recorded.

Some conversations happen too late.
This one happens now.

Be ready to listen.

#ninjacast#PQCapocalypsenow
View on LinkedIn ↗
ScottG

Someone mentioned "Agentic MDR" as the next biggie, and I got triggered...

View on LinkedIn ↗
ScottG

Only active really since August - god help you all next year 🤡🚀👾!

Happy new year 🎆🎈

View on LinkedIn ↗
The Probably Fine Daily

Threat intelligence every morning — new victims, new groups, what matters, in plain English. Free, with receipts.

Subscribe to the Daily →