# March 2026 Meetup Minutes

The meetup began with a quick round of introductions, followed by a
walkthrough
of the agenda. A balanced mix of students and professionals contributed
actively
to the discussions.

## Spec-Driven Development by Revanth A

Revanth began his talk with an introduction to his domain, automotive
software and how AI has been adopted in software development within his
organization. He explained the key problems addressed by spec-driven
development
using AI and described the combination of AI models used in his prompt
engineering workflow.

He also touched upon agile methodologies used in organizations and the
influence
of AI on these processes. Among various AI-driven frameworks, he
highlighted the
**BMad method**. He shared how he explored this approach to build an Android
application using Kotlin, a language he was not previously familiar with.

He demonstrated a screen recording showing how he selected options and used
the
BMad method to generate a PRD (Product Requirements Document) using Cursor
IDE.
He also emphasized the importance of checkpoints, where development can be
paused and resumed without needing to rework prompt engineering once the
PRD is
generated.

## Lightning Talk: OpenSSH Backdoor by Madhav Kumar

Madhav Kumar began his talk with a compelling introduction to the flow of
the XZ
backdoor, from XZ to systemd to OpenSSH. He discussed the widely known
malicious
activity involving a developer named “Jia Tan,” who contributed to the XZ
compression library originally maintained by Lasse Collin.

He explained what OpenSSH is and how it interacts with XZ. Furthermore, he
described how the backdoor was implemented using techniques involving the
Global
Offset Table (GOT), IFUNC, and other low-level mechanisms to enable
privilege
escalation.

He further elaborated on how the malicious code was designed to blend into
normal execution flow, and how a vulnerable version of XZ was included in a
release candidate of RHEL 10.

Madhav walked through the timeline of Jia Tan’s involvement in the XZ
project,
leading up to the discovery and resolution of the issue. He highlighted how
Andres Freund, a German engineer (not primarily a security expert),
identified
the anomaly due to an unexpected delay of over 500ms during SSH connections.

## A Software Development Process for the AI Era by Abishek

Abishek opened his talk by discussing how AI workflows have become the de
facto
standard in both individual and organizational software development
processes.
He presented a simplified view of the Software Development Life Cycle
(SDLC),
integrating both AI-driven and traditional approaches.

He explained how more conservative workflows are required in high-stakes
domains
such as banking, where human intervention remains critical. He referenced
the
DORA report, backed by Google Cloud, and described how AI acts as an
amplifier, exposing existing gaps in SDLC processes.

He also discussed the risks associated with continuous deployment in
AI-assisted
environments. Based on insights from the DORA report, he proposed
improvements
to development pipelines, identifying areas where human intervention is
essential versus where AI can take the lead.

He concluded by emphasizing that while specification and coding can largely
be
AI-driven, verification should remain human-led, supported by AI tools,
acting
as a release gate. He ended with open questions that led to engaging
audience
discussions.

## Networking Tea Break

## Overview of the Linux Graphics Stack by Praveen

Praveen shared his journey into understanding the Linux graphics stack,
which
began while attempting to install NVIDIA graphics drivers, a process he
found
challenging.

He explained how Linux systems follow a client-server model and how this
paradigm differs in the context of graphical systems like X11. Drawing from
his
web development background, he highlighted these differences in perspective.

He described the graphics pipeline, explaining how user input events
propagate
through the system to produce output on the screen. He compared X11, a
legacy
windowing system, with Wayland, a modern alternative.

He also introduced kernel-level components such as fbdev and DRM (Direct
Rendering Manager). To conclude, he demonstrated how Linux’s philosophy of
“everything is a file” can be leveraged to directly manipulate display
pixels
via fbdev.

## Lightning Talk: Neural Networks by Arun

Arun introduced the fundamentals of neural networks, explaining how
interconnected layers process information. He described a basic structure
consisting of an input layer, multiple hidden layers, and an output layer.

He drew parallels between artificial neurons and biological neurons,
emphasizing
how weights and biases influence outputs. He also demonstrated how these
parameters can be visualized to better understand neural network behavior.

## Lightning Talk: Product-Based Thinking by Arockiaraj

Arockiaraj shared his experience building a transaction insights dashboard
using
Google Scripts by parsing email based transaction notifications.

He explained how he leveraged Gemini AI and other tools to develop an
Android
application for expense tracking. He also showcased his embedded e-ink
display
device, which displays expense data on his desk to encourage mindful
spending.

He concluded by emphasizing the importance of product thinking, starting
from
identifying real world problems to building effective solutions. He
highlighted
how AI-assisted tools have made implementation significantly more
accessible.

-- 
Regards,
BabuSubashChandar
_______________________________________________
Chennaipy mailing list -- [email protected]
To unsubscribe send an email to [email protected]
https://mail.python.org/mailman3//lists/chennaipy.python.org
Member address: [email protected]

Reply via email to