Telegram macOS Developer Says 99.9% of Client Code Is Now Written by AI and Suggests Firing Manual Coders
Telegram macOS developer Mike has claimed that generative artificial intelligence now produces 99.9 percent of the code for the messenger's official clients. He proposed that developers who insist on writing code manually should be fired.
In a public discussion inside Telegram, Mike presented several examples of rapid AI-assisted development. One project, the website gram.org, was reportedly built with AI tools in a single day; Pavel Durov provided the domain after approving the idea. Another case involved adapting the iOS client for a foldable iPhone Duo device overnight using the Claude model.
Mike noted that designers within the team are now directly contributing to client code thanks to AI assistance. He added that between 80 and 90 percent of Telegram developers share his assessment of the technology's value.
Participants in the discussion questioned the quality and security implications of AI-generated code. Mike responded that vulnerabilities are also detected and remediated by AI, resulting in what he described as higher-quality output than traditionally written code. He characterized the workflow as a closed loop in which one neural network writes the code and another verifies it.
No comparative security audit results were provided to support these claims. Mike also announced that a new GitHub repository containing the macOS client source code will be opened, allowing external contributors to submit changes. The previous repository had remained inactive for more than a year.
Related articles
Rostelecom Suffers Mass Outage Disrupting Website, Mobile App and Personal Account
Rostelecom experienced a significant service disruption that left its official website, mobile application, and customer personal account either unavailable or functioning with errors. The first reports of problems began arriving around 12:40 Moscow time. Over a 24-hour period, the monitoring service Detector404 recorded approximately 2,500 user complaints. The largest share, 59 percent, originated from the Nizhny Novgorod region, followed by 4 percent from Kursk region and 2 percent each from Ryazan and Ivanovo regions. Rostelecom has not issued any official statement regarding the incident, and the root cause along with the timeline for full restoration remain unknown. The outage occurred in the same week that hosting provider ESTT suffered a similar infrastructure failure.
Russian State Duma Rejects Bill Proposing Fines for Covering Face in Public
On November 11 the State Duma rejected bill No. 1159193-8 that sought to introduce a new article in the Code of Administrative Offenses punishing appearance in public places with a face covered in a way that prevents identification. The draft included exceptions for medical reasons, official duties, weather conditions, and sports or cultural events. Two committees of the Federation Council on defense and security as well as constitutional legislation opposed the measure. The Duma committee on criminal and administrative legislation found the wording too vague and lacking legal certainty. It also noted that new offenses require prior federal-law prohibitions. On the same day lawmakers rejected another bill introducing criminal liability for crimes committed with deepfakes.
Basis Launches Basis AIR Platform to Optimize Shared GPU Resources for AI Projects
Basis, part of the RTK Data Centers group, has introduced Basis AIR, a platform designed to allocate computing resources across multiple AI initiatives within organizations. The solution targets small, medium, and large enterprises that have invested in expensive GPU hardware but face uneven utilization, with some teams waiting for capacity while others leave accelerators idle. According to the company, Basis AIR can increase accelerator utilization from 30% to 80% by enforcing task priorities, departmental quotas, and dynamic resource sharing. The platform supports running multiple workloads on a single accelerator and allows teams to borrow additional capacity when resources are available. It generates detailed consumption reports by project and division, helping executives decide whether new GPU purchases are truly necessary before approving further hardware acquisitions. Basis AIR is deployed on-premises in the customer’s data center, operates without internet connectivity, and integrates with third-party infrastructure platforms without requiring other Basis products.
30 Tools for Six Core Information Security Tasks: From Recon to Forensics
The article provides a practical guide for system administrators managing a single website, database, and server setup. It covers 30 specific tools across six categories: reconnaissance, web application testing, infrastructure scanning, incident response, digital forensics, and secure development practices. Each section recommends five tools with explanations of their use cases, such as discovering forgotten assets like test2 or analyzing network traffic during off-hours incidents. Tools include Subfinder for passive subdomain enumeration, Burp Suite for request inspection, Nmap for port scanning, Zeek for structured logging, Autopsy for disk image analysis, and Gitleaks for secret detection in Git history. The guide emphasizes proper documentation of assets, including ownership confirmation by colleagues like Lena, and highlights limitations such as VPN effects on scan results or TLS termination at load balancers. It concludes by stressing the value of maintaining accurate records to simplify future certificate renewals and incident handling.