Lakshan Weerasinghe

Lakshan Weerasinghe Daily tech news, practical tips, and clear insights into modern technology

The 2025 Cloud Security Reality Check: Critical Insights Every Leader NeedsAre we winning the war on cloud security? New...
17/02/2026

The 2025 Cloud Security Reality Check: Critical Insights Every Leader Needs

Are we winning the war on cloud security? New data suggests the battleground is shifting faster than defenses can keep up.

Cloud adoption has ceased to be a "strategy" it is now the standard. However, as organizations race to scale, security often trails behind. I recently reviewed SentinelOne’s latest 2025 Cloud Security resources, and the findings are a wake-up call for CISOs, DevOps engineers, and security architects alike.

Drawing from the 2025 Cloud Security Survey Report and the 2025 Cloud Verified Exploit Paths and Secrets Scanning Threat Report, here is a breakdown of why the threat landscape has changed and what you need to do about it.

1. The Visibility Gap is Still Widening

According to the survey, which gathered insights from over 400 security professionals, the fundamental struggles remain surprisingly consistent. Despite the abundance of tools available, organizations are still battling the "Big Three" of cloud chaos:

Misconfigurations: The leading cause of breaches.

Poor Visibility: You can’t protect what you can’t see.

Fragmented Defenses: Siloed tools create blind spots.

Perhaps most concerning is that the Public Cloud is now viewed by many leaders as the riskiest environment in their infrastructure. As complexity grows, the confidence in securing these environments seems to be wavering.

2. The "Secrets" Aren't So Secret

Theoretical risks are one thing; real-world data is another. SentinelOne’s threat report analyzed thousands of customer environments and uncovered a staggering statistic: 100,000+ exposed secrets.

We are talking about:

API Keys left in open repositories.

Hardcoded credentials buried in container images.

Default access tokens that were never rotated.

Attackers are no longer just looking for software vulnerabilities (CVEs); they are hunting for these keys. The report highlights how attackers utilize Verified Exploit Paths specific, proven routes to breach cloud workloads using these exposed secrets.

3. The New Wave of Threats: AI and Shadow IT

The threat landscape isn't static. The 2025 outlook identifies several accelerating vectors that are keeping security teams up at night:

Shadow Deployments: Infrastructure spun up without IT oversight, leaving it unpatched and unmonitored.

Hardcoded Secrets in Containers: Developers prioritizing speed over security, baking keys directly into code.

AI Tool Leaks: As teams rush to adopt AI tools, sensitive corporate data is increasingly being fed into public models or exposed via insecure AI integrations.

4. Moving Beyond "Theoretical" Security

The traditional approach of scanning for every single CVE produces too much noise. Security teams are drowning in alerts for vulnerabilities that aren't actually reachable or exploitable in their specific environment.

This is where SentinelOne’s Singularity Cloud Native Security (CNAPP) distinguishes itself.

Instead of flagging everything, it focuses on context.

Agentless Scanning: rapid visibility without the friction of deploying agents everywhere.

Offensive Security Engine: It thinks like an attacker.

Autonomous Prioritization: It differentiates between a "theoretical" risk and an actual exploitable path.

By focusing on what can actually be exploited, teams can stop chasing ghosts and start fixing real holes.

The Bottom Line
If you are responsible for cloud security, hoping for the best is not a strategy. The 2025 reports from SentinelOne are essential reading to understand the specific mechanics of modern breaches.

Just saw this interesting report from Anthropic about their latest AI model, Claude Opus 4.6. They did a deep dive into ...
11/02/2026

Just saw this interesting report from Anthropic about their latest AI model, Claude Opus 4.6. They did a deep dive into "sabotage risks" (like whether it could secretly mess things up or help with really bad stuff). Overall, they say the risk is very low (but not zero), and there are some concerning bits in tests like small help with dangerous things when pushed hard.

Crazy how fast AI is moving and how seriously they're taking safety now. What do you guys think? Worth reading if you're into AI stuff

Link in comments 👇

09/02/2026

Free Cloud Service for Backup Your Data 😍

AI is the  #1 Security Threat for US Banks & Financial Firms in 2025Did you know? 59% of financial services leaders say ...
09/02/2026

AI is the #1 Security Threat for US Banks & Financial Firms in 2025

Did you know?

59% of financial services leaders say the fast-moving AI ecosystem is their TOP security concern right now even higher than data breaches or ransomware.

Key stats from the brand-new 2025 Thales Data Threat Report (Financial Services Edition):

👉57% worried about AI model & data integrity
👉Only 15% of organizations encrypt 80%+ of their sensitive cloud data
👉SaaS apps jumped 27% to 107 on average more doors for attackers
👉Quantum computing threats? 57% already fear future encryption breaks

US banks, insurers, and fintech's are adopting AI faster than anyone… but are we ready for the risks?

Download the full report for free and see how to protect your data in the AI + Quantum era (Link in Comment)

What’s your biggest worry? GenAI risks or quantum threats? Drop a comment below.

The dream of “autonomous coding” is closer than ever, but as any seasoned engineer knows, the devil is in the deployment...
07/02/2026

The dream of “autonomous coding” is closer than ever, but as any seasoned engineer knows, the devil is in the deployment.

Recently, I decided to put some of the most advanced AI tools to the test Antigravity Pro (using Claude 4.5 Thinking/Opus) and Gemini 3 Pro. I wanted to see if an AI agent could navigate a complex, real-world SaaS environment without handholding.

The results were impressive, chaotic, and a massive wake-up call for the future of our profession.

The Setup: A Multi-Headed SaaS Beast
The project wasn’t a simple “To-Do” app. It’s a complex SaaS ecosystem featuring:

One Backend powering five different Frontend systems (Project, Compliance, Finance Management, etc.).
AWS CDK Infrastructure: Meaning the backend cannot be run locally; it must be deployed to the cloud for testing.
Strict Workflows: Technical docs must be approved by Seniors, PMs, and DevOps before a single line of code is written.
I gave the AI a User Story for a large feature in the Compliance Management module and a very detailed prompt. However, I intentionally withheld four “human” insights about our specific environment:

The inability to test locally.
Deployment is only allowed from the dev-temp branch.
The Service Stack file was at its limit (needing a new stack).
A “Double-Deploy” requirement to prevent CORS issues with AWS Cognito.

The Good: Coding Standards and UI Precision
The AI started strong. It analyzed the database schema, updated the scripts, and built the backend following our existing patterns (Dependency Injection, RESTful APIs, etc.).

When it moved to the frontend, I was genuinely impressed. Using only Figma screenshots, it built a beautiful, responsive UI that matched our design system perfectly. It refactored mock data into real API calls using Axios without breaking a sweat. 48 files were modified or created, and the code quality was top tier.

The Bad: When the “Agent” Lost the Plot

The friction started during the deployment phase. This is where the difference between a coder and an engineer became clear:

Branch Blindness: It tried to deploy from my local dev-lakshan branch. After five minutes of "thinking" through errors, it correctly identified that it needed to switch to dev-temp.
The Resource Limit: It hit the AWS Service Stack limit. To its credit, it eventually figured out it needed to create a new stack.

The CORS Nightmare: Because it didn’t know about the “double-deploy” rule for Cognito, it hit CORS errors during testing. Instead of questioning the infrastructure, the AI went down a rabbit hole altering pre-configuration files and trying to “bypass” AWS security.
The Warning Sign: If I had blindly approved those AWS changes, the entire project infrastructure would have been compromised. The AI began “hallucinating” infrastructure fixes that were actually destructive.

The “Human” Fix

I eventually had to intervene. I updated the prompt: “Once everything is complete, deploy the backend twice.” Suddenly, everything worked. Even the Pull Request (PR) process needed a human touch; the AI initially tried to merge dev-temp directly into the main/dev branch instead of returning to the feature branch first.

Key Takeaways for the AI Era

UI is AI’s Strong Suit: Frontend development and design-to-code translations are becoming incredibly efficient.
Infrastructure is the Final Frontier: AI still struggles with “tribal knowledge” those specific quirks of a company’s AWS/DevOps setup that aren’t written in the documentation.

The “Senior” Filter: You cannot give these tools to someone who doesn’t understand the foundation. An inexperienced dev would have pushed the “CORS bypass” and broken the production environment.

Final Thoughts: Will SE Jobs Vanish?
No, but they are changing. The “Coder” who just translates requirements to syntax is at risk. However, the Software Engineer the one with the Basic + Foundation + Skills + Experience will become a “Pilot.”

AI will do the heavy lifting, but we are the ones who know when the plane is flying into a storm. To survive the AI wave, don’t just learn to code; learn how the whole machine works.

This is an absolutely valuable tool 😍Especially for interns or someone who has just joined a team, understanding a large...
09/01/2026

This is an absolutely valuable tool 😍

Especially for interns or someone who has just joined a team, understanding a large code repository can take quite a bit of time. Usually, a senior developer or another team member has to spend time giving KT (knowledge transfer), which can be tiring and inefficient.

This new tool introduced by Google solves that problem. You simply provide the repository URL, and it automatically generates a visual and well-structured documentation, including:

Architecture overview

API references

Class explanations

Function descriptions

In other words, it creates Architecture Diagrams, Class Diagrams, and Sequence Diagrams for you.

The most important part is that whenever the repository gets updated (for example, when a PR is merged), the documentation is automatically updated as well. This completely eliminates the issue of outdated documentation.

We’ve all experienced this when joining a new job sometimes the provided docs are old, while the code has already changed. That makes understanding the system really difficult.
This tool directly solves that problem.

On top of that, if there’s any confusing part in the codebase, you can open the built-in chat option and ask questions. It will explain those unclear sections in a simple and understandable way.

Currently, this tool is released as a public preview and works only with public repositories. Support for private repositories is marked as coming soon on the website.

Give it a try and see for yourself.





Address

Ratnapura

Opening Hours

Monday 09:00 - 17:00
Tuesday 09:00 - 17:00
Wednesday 09:00 - 17:00
Thursday 09:00 - 17:00
Friday 09:00 - 17:00
Saturday 09:00 - 17:00
Sunday 09:00 - 17:00

Alerts

Be the first to know and let us send you an email when Lakshan Weerasinghe posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Shortcuts

Share