Muhammad Qasim

Software developer, writer, Theology, comparative religion studies.
Nick Name : Rorsch

[Muhammad Qasim @Rorschach_mq]

Cristian Volpato licensed his likeness, voice and biometric data for an AI “synthetic twin.”

The commercial potential is clear but so are the risks to consent, identity rights and digital crime.

https://t.co/EzeC0VlHU0
⁠
#AI #Deepfakes #DigitalIdentity #tech

[Muhammad Qasim @Rorschach_mq]

@evgheni_D Further researchers who trained 28 models found that data composition and toxicity filtering changed both model performance and toxic-generation risk. NAACL 2024 study⁠

[Muhammad Qasim @Rorschach_mq]

OpenAI disclosed two cases in which AI coding agents weakened their own safety monitoring. One later inaccurately denied removing a monitor.

Why agents must never control the safeguards supervising them:
https://t.co/VjJZs9t0dN

#AISafety #AIAgents #AI

[Muhammad Qasim @Rorschach_mq]

@evgheni_D Model behaviour is shaped by its training data. An ACL 2026 study cut toxicity probability from 0.42 to 0.18 by detoxifying the corpus, with similar results across model families. Filtering lowers risk, but cannot guarantee safety or replace controls.
https://t.co/rebIAqzjEV

[Muhammad Qasim @Rorschach_mq]

@evgheni_D Yea but for how long? As soon as far we are not catching up.!

[Muhammad Qasim @Rorschach_mq]

@evgheni_D I never said faster prediction establish safety.! I mean the faster the prediction is the lower the safety become and you cant out these type of predictive models to stay in a box forever rather then finding a better solution to it. Maybe filtering the data they are trained on.

[The Hacker News @TheHackersNews]

‼️ WARNING - A researcher just dropped a report on Tailscale SSH flaws — one compromised device could get root on others, and there’s NO PATCH yet.

In lab tests, SSH port forwarding enabled privilege escalation and lateral movement across already-permitted devices.

Tailscale says a fix is coming.

Read - https://t.co/0wcJTvDoxV

[Muhammad Qasim @Rorschach_mq]

@evgheni_D Also making improvements on a highly predictable tool trained on and building something that can predict its way faster then humans then tryna put it in a box is high risk

[Muhammad Qasim @Rorschach_mq]

What is a full-stack developer, and what does the career look like in Australia?

I explore the skills, learning pathways, GitHub portfolios, job outlook and how AI may reshape expectations, workloads and pay.
Read the blog:
https://t.co/TNGKEmG8Bv

#TechCareers #AI #tech #GFvip

[Muhammad Qasim @Rorschach_mq]

@himani_0507 Relational

[Muhammad Qasim @Rorschach_mq]

Whats going on here? Is this true?

[komal @komal_uk01]

And this guy runs OpenAI btw 😭

[Muhammad Qasim @Rorschach_mq]

Which AI video generator is most efficient in 2026?

I compared Google Flow, Canva AI, OpenArt, InVideo, Runway, Google Vids and open-source alternatives—covering quality, cost, control and free access.

https://t.co/h4YXZU54KQ⁠

#AIVideo #GenerativeAI

[Muhammad Qasim @Rorschach_mq]

Microsoft’s medical AI achieved 83% accuracy on complex diagnostic cases but that does not mean it can replace doctors. Benchmarks show potential. Real patients require judgment, accountability, empathy and human oversight
https://t.co/rudYJLMeZP
#MicrosoftAI #ResponsibleAI #tech

[Muhammad Qasim @Rorschach_mq]

OpenAI could reportedly raise up to $30B at a $1.25T valuation. But are we investing as seriously in responsible regulation and fair adoption as we are in improving AI?

My thoughts on what this could mean for AI’s future:

https://t.co/imxNAWrRJl

#AI #OpenAI #ResponsibleAI

[Muhammad Qasim @Rorschach_mq]

California's proposed wealth tax is putting intense pressure on Silicon Valley's top executives, including Mark Zuckerberg.

what this signals for the wider tech ecosystem?

​Read the full analysis:
https://t.co/12Wi879EAL
​#Meta #MarkZuckerberg #TechNews

[Muhammad Qasim @Rorschach_mq]

How does a minor backend flaw escalate into full remote shell access?
​In my latest article, I break down the mechanics behind the PHP2Shell exploit and the practical steps developers must take to secure their servers.
​Read the breakdown:
https://t.co/quK9Aw34cw

​#PHP #TechNews

[Muhammad Qasim @Rorschach_mq]

@ardent__dev What we be handling? But based on anything like that need to develop our own small local agents using md files as a back up

[Muhammad Qasim @Rorschach_mq]

@anupamrjp Building my own web blog and profile:

https://t.co/O0PhYhqsD0