01 / The New Normal

Zero-days were the work of months. Now they are the work of minutes.

Zero-days were once rare. They were the product of months of manual effort, elite skill, and patient research. Mythos proves that is no longer true.

Mediocre attackers can now operate at elite levels. Discovery is faster than ever. Exploitation is faster than disclosure. The gap between the two, the gap that every modern security program is built around, is collapsing in real time.

"It will not be long before such capabilities proliferate, potentially beyond actors who are committed to deploying them safely."
Anthropic · Project Glasswing announcement
From months of expert effort to minutes of model output
MONTHS
Nation-state era
2010 to 2017
WEEKS
Disclosure to PoC
2017 to 2021
DAYS
Modern CVE cycle
2021 to 2024
HOURS
AI-assisted exploit dev
2024 to 2025
MINUTES
Mythos-class capability
2026 →

This is the world security tools have to operate in now. Not the one they were designed for.

02 / Oligo's Position

Every assumption needs to be reconsidered. Especially the ones built on patching.

The dominant model has been scan, prioritize, patch. It was already strained. AI-enabled attackers have broken it.

You cannot win a race when your opponent moves at machine speed and you move at change-management speed. The math has changed. The strategy has to change with it.

"Assume you can't patch."
Anton Chuvakin · Office of the CISO, Google Cloud Security
Where the industry is landing
  • Reachability is necessary, not sufficient.
  • Detection without prevention is a backlog with a dashboard.
  • Runtime is the only timeline attackers cannot outpace.
Discovery is not the constraint
Scanners find vulnerabilities faster. Attackers exploit them faster. Discovery speed is not the bottleneck. The patch cycle is.
Prioritization is still reaction
Reachability analysis, exploitability scoring, risk-ranked backlogs. All of it still leaves you reactive. You are deciding which fires to fight while new ones are being started.
Patch SLAs assume time you no longer have
The CVE-and-patch model assumes you can remediate before attackers act. That assumption no longer holds when exploitation timelines collapse from weeks to hours to minutes.
Supply chain risk has no CVE
Malicious code can be introduced upstream without ever getting a CVE. Traditional detection has no hook to catch it until execution.

The technical reality

Runtime protection has been promised before. Previous attempts failed.

RASP saw too much and understood too little. Latency, false positives, crashed applications. Most teams turned it off within months.

Cloud security vendors added response capabilities, but kill-the-container is a blunt instrument. It cannot see inside the application. It cannot know what code is running. It cannot tell when that code starts doing something it should not.

The result: runtime blocking exists in product demos. Almost no one runs it in production. The risk of breaking the app outweighs the risk of the vulnerability.

The missing piece was always context.

Scan · Prioritize · Patch

Backlog grows faster than you remediate.

OPEN HIGH/CRITICAL 0
Runtime Block

One technique blocked. Application uninterrupted.

EXPLOIT
KERNEL · BLOCK POINT
SYSCALL BLOCKED
APPLICATION
HEALTHY
03 / How Oligo Holds Up at AI Speed

Oligo was built to defend against attackers that move at AI speed.

We were founded in 2022, when everyone else was pushing agentless. We knew the world was heading somewhere static scanners couldn't follow. That world has arrived.

A · DEEP APPLICATION INSPECTION

See running code at the function and syscall level. In production. Without source access.

Oligo's runtime sensor captures execution telemetry at the function and syscall level, in real time, in production, without requiring source code access. By observing code as it runs, Oligo can tell the difference between what an application is supposed to do and what an attacker is trying to make it do. Static scanners, network tools, and infrastructure-layer controls cannot make that distinction.

OLIGO PRODUCT Profile Violation · function-level evidence
Oligo product UI showing a function-level call graph with a forbidden call flagged, leading to a spawned process. This is the Profile Violation evidence surfaced by the Oligo sensor.
A real Profile Violation in the Oligo console. The sensor reconstructs the application's call graph in production and flags the exact function call that crossed the line into the runtime exploit chain.
B · RUNTIME EXPLOIT BLOCKING

Stop exploits inside the application. Without disrupting uptime.

Oligo correlates application-layer function calls with OS-level syscalls. A specific sequence indicates an exploit attempting to execute. Neither element is malicious in isolation. The combination is. When Oligo detects the pattern, it blocks the syscall directly from the kernel. The attack stops. The application keeps running.

Application call stack
1flask.app.Flask.dispatch_request
2myapp.routes.calculate
3myapp.logic.evaluate_expression
4builtins.eval
5socket.create_connection
6socket.socket.connect
OLIGO
SENSOR
OS-level syscall
connect( 10.0.0.5:4444 )
Outbound network call. Reverse shell signature.
TECHNIQUEEval Injection
CWECWE-95
MONITORING...
Technique-based
Block the technique once. Neutralize every CVE that uses it. Including future ones.
Zero added latency
Enforcement happens directly in the kernel. No proxies. No instrumentation overhead.
Surgical precision
Oligo blocks the syscall, not the container. Traffic continues normally.
Code you don't own
No source access required. Protect first-party apps and hosted third-party software alike.
C · THE SCALE ADVANTAGE

Protection scales. Patching doesn't.

A single Oligo blocking rule for insecure deserialization covers more than 100 HIGH and CRITICAL Java CVEs and 80+ Python CVEs in an average cloud environment. One rule. When the next deserialization vulnerability is disclosed next month, organizations running Oligo are already protected.

Technique-based protection closes entire attack surface classes rather than chasing individual vulnerabilities. It is the only model that can keep pace with AI-accelerated discovery.

And the patch backlog you still have

Oligo uses Deep Application Inspection to identify which vulnerable components actually execute, highlight the vulnerabilities that are reachable in real flows, and de-prioritize the noise. Teams patch the risk that is real in their environment, instead of patching everything in alphabetical order.