An organization that offers Cyber Security news & services, Information Security training, threat and vulnerability Assessment, web-app development.© 2019-2026.
Joined December 2019
- Tweets1.7K
- Following0
- Followers268
- Likes340
Pinned Tweet
OUR JOURNEY FOR 2026:
- The goal is to bring knowledge in a different form, from what is familiar.
- With great insight on terminologies, its theoretical and practical applications.
- You don't just learn from our courses, but also infused with knowledge on how to connect the dots.
#FixitgearwareSecurity #RoadMap2026
Dear @Microsoft Teams chat is amazing and good for work related meetings & collaboration.
As a heavy and concerned user, can you create a tab called “Group Chats” ? To enhance productivity, and not cluster messages of single chats together with group chats created.
“Unread —> Chats —> Meeting chats —> Group chats”
Thanks
Concerned Users @fixitgearware
#Collaboration #Microsoft #Teamschat
Windows OS Update:
@Microsoft has implemented its new Phone-Link feature on the Windows OS.
Users across various countries and regions, should be expecting this feature gradually rolled out on their Windows Operating systems.
#FixitgearwareSecurity #TechInformations #Updates
The Phonelink feature, allows the user to then interact with their mobile phones on the paired computer.
With a cool UI, the paired hardware functions is then accessible over the workstation it is paired with.
Some of the accessibilities includes:
1. Ability to make calls directly from the connected computer.
2. Read messages and messages in other applications installed on the paired mobile via the computer.
3. with even more capabilities to explore.
Have the feature enabled on your computing device? Test it out, and share what you like about it, in the comment.
#FixitgearwareSecurity #TechUpdates
Check out this week’s content as we connect our custom miniature agent to carry out some quick recon activities, while still facing the guardrails limitations of Gemini Argon4.
Link:
youtu.be/bJfmmey4JT4
#Fixitgearware #Security #LLM
Video will be up in about 70-mins 😎.
You get to See how GeminiArgon4 failed with its guard rails, and how Codex daybreak blue yield valuable ready targets subdomains for vulnerability research.
#AgenticRecon #PenetrationTesting
Demonstrating with these two Cyber based models, to ascertain their recon strengths and weaknesses.
Building a powerful agent, is all about leveraging the strengths of a specific model and wrapping your agent construct around it.
What do we mean ?
The truth most tech leads and orgs will not expose when talking about how capable a model is, is based on what context they were tested and benched marked upon.
While LLM is useful, some are mortals, and some over time will be higher mortals, just like we have intelligent beings, geniuses, and naturals.
However, in the context of LLMs, data-set upon which they were trained is not just crucial, but a necessity to determine how powerful they will evolve to become more capable.
This week content focuses on live demonstration, by just testing our miniature recon-agent, with its unique workflow outline, and same prompts on two unique AI platforms of our choice.
1. @OpenAI Day-Break-Blue (Cyber).
2. @GoogleAI Argon-4.
The idea of these two selection is based on two criteria's.
A. @OpenAI (Being a pioneer in the LLM industry).
B. @Google (Being the master forefront of data-sets based on its years of search engine existence).
Both have fair trials based on these strengths we outlined, and that determines what we will do next with our Robust Recon Agent.
Video drops:
- 13:00 HRS US Time.
- 25-October-2026.
#Fixitgearware #Security #LLM #AgenticRecon #Testing #Miniatureagents
🤖 Made with AI
The most expensive Ai services / LLM out there is @GoogleAI
Apparently subscription to any of its paid package gives you access to only web-interface usage.
1. Pro subscription covers only web interface usage & workspace integration.
2. To have access to its latest model “Argon” via API calls despite subscription, requires a pay-as-you-go credit.
3. Usage of Vertex Ai/GCP needs additional pay-as-you-go credit as well.
At this point, @Google is making its Ai services overly expensive, compared to other competitors.
Breaking down the cost running a minimal token prompts might just require $100 cost per month over API. Keyword “minimal token prompts”.
This week we are going to go Recon Kamikaze demonstrating with two outstanding LLM wrapped around our custom recon Agent.
1. @OpenAI Codex Daybreak Blue.
2. Stay-tuned (on the next LLM of our choice to use).
The output of our perfect recon results would be compared over the two LLM of our choice, their similarities and differences with respect to recon time generating the desired results.
Video drops:
25th Oct 2026 | 13:00 HRS US Time 🕰️.
Location: YouTube.
Harness, Agents, and Agent. md files explained:
youtu.be/YAQ-HB2Rqjk
MCP Explained demonstration with shodan MCP server:
youtu.be/85sVwKay8f0
#Fixitgearware #Security #Ai #Openai #Codex #Daybreakblue #mysteryAi
This week we are going to go Recon Kamikaze demonstrating with two outstanding LLM wrapped around our custom recon Agent.
1. @OpenAI Codex Daybreak Blue.
2. Stay-tuned (on the next LLM of our choice to use).
The output of our perfect recon results would be compared over the two LLM of our choice, their similarities and differences with respect to recon time generating the desired results.
Video drops:
25th Oct 2026 | 13:00 HRS US Time 🕰️.
Location: YouTube.
Harness, Agents, and Agent. md files explained:
youtu.be/YAQ-HB2Rqjk
MCP Explained demonstration with shodan MCP server:
youtu.be/85sVwKay8f0
#Fixitgearware #Security #Ai #Openai #Codex #Daybreakblue #mysteryAi
The whole noise will be explained in October with consistent video release in series.
- Agents (first free video is on the demo of creating the agents file, what they are, and a practical example). - date 04-Oct-26
- MCP (subsequent announcement)
- LLM (subsequent announcement)
- Automation Testing (subsequent announcement)
- Prompt (subsequent announcement)
A few demonstration starting with @OpenAI gpt-daybreak-blue-latest aka GPT-5.6 Sol.
Token Maxing, caused by not properly aligning the flow of your AGENTS.md (Agents), making your LLM high on cost consumption instead of cutting. C-course plan flow, and why we have to postpone our Cybersecurity course release this year, cos the past decade has been rendered obsolete by the presence of Ai.
#FixitgearwareSecurity
A very long week in the Industry:
1. @official_php (Remote Code Execution).
2. @gitlab new CVSS 10.0.
3. @opencode getting compromised.
4. @Meta getting hit with a $25-Billion dollar law suite over data theft 😅.
These occurrences has not only justified our opinion on the huge gap in the industry, but also proved that even top organisations consider “Hacking” as cybersecurity.
- No regulatory compliance (Facebook).
- No consistent code auditing (GitLab, and PHP).
- No LLM-eval (OpenCode).
- No data management policies (how data is obtained, stored, or shared)- Facebook.
- No organisation procedural steps on business policies and how versioning with releases are controlled and continuously audited (The entire organisation’s).
Cybersecurity is not hacking alone, nor certification. Organisations have to start looking at what and who they employ, if we don’t want to really wake up one day, to find out that our own very Internet 🛜 has destroyed us.
#Fixitgearware #Security
There is an African proverb which says “Where the Honey 🍯 comb is, there you find the Honey Badger. 🦝”
The internet 🛜, was designed without considering security as the backbone of its foundation, and one 1️⃣ would think the Ai and agentic era could fix that, by designing Ai 🤖 with security as the very foundation before everything else.
However, that is not the case. Capitalism and wanting to be the first to have the highest thinking model is what all these organisations are focused on.
Well the customers and users of these services are the “Honey Comb 🧀”.
FixitGearware Security certainly would change the face of Cybersecurity, time is all we need, while perfecting on non-capitalist mindset. 😎
The past 72-hrs we have been conducting a comprehensive investigation on a series of hardwares built on LTSC images dated way back early 2000s.
Core Tools:
1. Copilot Ai.
2. Windows Powershell.
3. .evtx consolidators.
4. Microsoft Excel.
5. Decades and years of experience.
Emphasis on “No.5” made every findings possible.
At some point we were analysing app evtx logs to detect abnormal reboots instead of system based evtx (blame it on sleepless nights).
The catch?
Grabbed a screenshot and fed it to copilot request why specific event IDs are missing (years of experience).
Copilot: “I am not seeing any trace that the right logs is been reviewed”
We: “Check our log path nomenclature” Only to discover that we are reading from a different evidence folder.
Proceed to access the right captured evidence folder, boom all “Event-ID” found.
Our take:
1. Ai makes things faster and easier.
2. Years of experience are required to use Ai faster and get accurate results without hallucinations of the Model.
3. Paid tools are most times not necessary.
4. You always need some good sleep. 😴 😂😂.
#Fixitgearwaresecurity #Forensics #BlueTeam #Talks #Copilot #Microsoft #Ai
A 75-year old woman was found to have defrauded financial institutions to a tune of 30-million USD, within a span of 10-months (July-2017 to May 2018).
FBI has declared a bounty for any useful information about her whereabouts.
Bounty: $150,000 (USD).
Source:
globalnews.ca/news/12105511/…
Next training will be on LLM.
Video Drop Date : 25-October-2026
Time: 11:00 A.M US timing.
Let’s test more of the best of models and provide unbiased analysis.
See You!
The whole noise will be explained in October with consistent video release in series.
- Agents (first free video is on the demo of creating the agents file, what they are, and a practical example). - date 04-Oct-26
- MCP (subsequent announcement)
- LLM (subsequent announcement)
- Automation Testing (subsequent announcement)
- Prompt (subsequent announcement)
A few demonstration starting with @OpenAI gpt-daybreak-blue-latest aka GPT-5.6 Sol.
Token Maxing, caused by not properly aligning the flow of your AGENTS.md (Agents), making your LLM high on cost consumption instead of cutting. C-course plan flow, and why we have to postpone our Cybersecurity course release this year, cos the past decade has been rendered obsolete by the presence of Ai.
#FixitgearwareSecurity
Patch Tuesday: KB5127648
@Microsoft has just released a patch for Windows 11-24H2: 25H2: 26H2.
The released update is said to upgrade systems running windows 11, to builds 26100.9711 for 24H2, 26200.9711 for 25H2, and 26300.9711 for 26H2.
The October update is said to fix the following related bugs:
- black screen issues and explorer crashes in multi-user environment such as Azure Virtual Desktop & FSLogix.
- VPN connection issues impacting Always On VPN connection failures.
Other related fixes includes:
- Fixed storage Spaces Direct virtual disk failing to come online after previous security updates.
- Issues impacting secure boot certificates updates and latest security vulnerabilities.
The said update also fixes playback issues impacting the AC-3 audio.
The core reasons Ai model hallucinates is based on the Limitations on the training data sets. @GoogleAI has the potential to build the most powerful Ai tool, cos the internet is Google, and Google is the internet.
Literally everything on the web is indexed by Google. Everything!!! 😎
This week’s content explains MCP, and with this explanation you will figure out if really, you need to always connect to MCPs of your choice or not, and when to do so.
youtu.be/85sVwKay8f0
To appreciate the content more, first watch our video on Agents, and agents-md files integration with LLM.
🚀 New video is live!
Learn how **Harness Agents, AGENTS.md, and Daybreak Blue** come together to create powerful **agentic workflows**.
🎥 YouTube:
youtu.be/YAQ-HB2Rqjk
#HarnessAgents #AgenticAI