Best of AI — September 2025

1
Article
Anton Zhiyanov·33w
Write the damn code
Software engineers should avoid endless prompt refinement when working with AI and instead actively write code themselves. Rather than trying to perfect prompts to get AI to generate complete solutions, developers should engage with the code directly by refactoring AI-generated code, writing initial versions for AI review, handling critical parts manually, or creating code outlines for AI to complete. This hands-on approach produces better results than attempting to 'program in English' through prompt iteration.
1.6K
103
2
Article
Josh M.·35w
GPT-5 is Trash.
ChatGPT-5 has received significant criticism from users who report that responses are shorter, blander, and less engaging than previous versions. Despite being marketed as PhD-level intelligence, the model still makes basic errors in math and reasoning while suffering from hallucinations. OpenAI's removal of model selection options and implementation of an autoswitcher has frustrated users, leading many to believe this was a cost-saving measure rather than genuine improvement. The backlash was severe enough that OpenAI restored access to older models like GPT-4o.
204
40
3
Article
Addy Osmani·33w
Give your AI eyes: Introducing Chrome DevTools MCP
Chrome DevTools MCP is a new tool that connects AI coding assistants to Chrome's DevTools through the Model Context Protocol, allowing AI agents to see, interact with, and debug live web applications in real browsers. The tool enables AI to perform tasks like running performance traces, inspecting DOM elements, monitoring network requests, simulating user interactions, and automatically fixing issues based on actual browser feedback rather than guessing.
198
4
4
Article
Modal·35w
Introducing Notebooks
Modal launches Notebooks, a collaborative cloud-based computing environment that provides instant GPU-enabled Python kernels starting in under 5 seconds. The platform offers real-time collaborative editing, automatic resource scaling from 0.125 CPUs to 8 H100/B200 GPUs, and seamless integration with Modal's existing infrastructure including Volumes and Functions. Key features include fast cold-start times, automatic idle shutdown to reduce costs, shared environments across teams, and modern development tools like LSP support and AI completions.
183
5
Article
Ayush Kumar·35w
Google's New Golang based framework for Building AI-powered apps
Google has released Genkit Go 1.0, the first stable and production-ready version of their open-source AI development framework specifically designed for the Go programming language ecosystem. This framework enables developers to build AI-powered applications using Go.
177
6
6
Video
Stefan Mischook·33w
In 2025, it is NOT about JS and React Development.
The software development landscape has shifted dramatically in 2025, moving away from traditional JavaScript and React skills toward AI-integrated development. Developers must adapt by learning to implement AI into workflows, build AI-first applications, and use AI tools to accelerate traditional coding. Entry-level positions are being automated, but opportunities exist for those who master AI development tools and can build solutions for non-tech companies seeking AI integration.
170
24
7
Article
Hacker News·34w
AI Was Supposed to Help Juniors Shine. Why Does It Mostly Make Seniors Stronger?
AI tools in software development are primarily benefiting senior developers rather than empowering juniors as initially expected. While AI excels at generating boilerplate code and automating repetitive tasks, it struggles with code review, architecture design, security considerations, and quality assessment. Senior developers can better leverage AI's strengths while avoiding its pitfalls, as they possess the experience to write effective prompts, evaluate AI-generated code, and maintain proper software architecture. The current state suggests AI is concentrating power among experts rather than democratizing coding skills.
122
29
8
Article
Hacker News·36w
MIT Study Finds Artificial Intelligence Use Reprograms the Brain, Leading to Cognitive Decline
MIT researchers conducted a study using EEG brain scans to examine how ChatGPT usage affects cognitive function during essay writing tasks. The study found that students who relied on AI assistants showed weakened neural connectivity, impaired memory recall, and reduced sense of ownership over their work. Participants using ChatGPT were unable to quote from essays they had just written, while those using traditional methods or search engines maintained better cognitive engagement. The research suggests that AI dependency leads to cognitive offloading, where the brain adapts to rely on external tools at the expense of critical thinking and memory formation.
119
19
9
Article
Visual Studio Blog·35w
Visual Studio 2026 Insiders is here!
Microsoft announces Visual Studio 2026 Insiders, featuring deeply integrated AI capabilities, significant performance improvements for enterprise-scale development, and a modern UI redesign using Fluent UI. The release introduces the new Insiders Channel replacing the Preview Channel, includes GitHub Copilot Free integration, and promises faster solution loading, building, and debugging across both x64 and Arm64 architectures. The IDE now offers enhanced code understanding, automated testing suggestions, performance profiling with AI-powered fixes, and 11 new color themes.
106
14
10
Video
The Coding Sloth·35w
AI Makes Programming Fun
Programming with AI tools can make development more enjoyable by automating boring tasks like learning new codebases, writing PR summaries, and handling code reviews. The author demonstrates using AI to contribute to VS Code's open source project, showing how AI can help understand large codebases, generate code, and streamline workflows. While AI tools like code completion and automated agents can speed up development and reduce tedious work, they're not perfect and require careful use to avoid over-reliance and potential errors.
101
3
11
Article
freeCodeCamp·36w
How to Fine-Tune Large Language Models
A comprehensive course covering fine-tuning techniques for large language models, including supervised fine-tuning, reinforcement learning with human feedback (RLHF), and QLoRA methodology. The course explains the differences between fine-tuning, pre-training, and prompt engineering, with practical applications and case studies for specializing LLMs for specific domains.
92
3
12
Article
Chrome Developers·33w
Chrome DevTools (MCP) for your AI agent
Chrome DevTools now integrates with AI coding assistants through the Model Context Protocol (MCP), allowing AI agents to debug web pages directly in Chrome. This enables AI assistants to see what their generated code actually does in the browser, analyze performance traces, diagnose network errors, simulate user behavior, and debug styling issues in real-time. The public preview is available via npm and aims to improve AI coding accuracy by providing browser debugging capabilities.
91
1
13
Article
controversy.dev·33w
Is vibe coding just a "workslop" factory?
A BetterUp Labs and Stanford study reveals that 40% of employees have encountered 'workslop' - AI-generated content that appears professional but lacks substance to advance meaningful work. The research highlights growing concerns about AI tools producing low-quality output that masquerades as productive work.
89
30
14
Article
Medium·33w
7 AI Coding Techniques That Could Save You 10+ Hours This Week
Seven practical AI techniques for developers to automate repetitive coding tasks including component generation, API scaffolding, debugging assistance, code refactoring, UI creation, test generation, and deployment automation. Each technique includes structured prompts, real examples, and estimated time savings ranging from 45 minutes to several hours per task.
88
7
15
Article
UX Planet·33w
7 UX Skills that will be DEAD by 2026: AI will replace them
Seven traditional UX design skills are predicted to become obsolete by 2026 due to AI automation and changing industry demands. Static wireframing in Figma is being replaced by interactive prototypes built with AI-powered tools like Replit and Lovable. Manual handoffs to developers are giving way to design-to-code workflows that generate functional components. Traditional usability testing is being supplemented by AI-driven analysis and hybrid testing systems. The shift moves away from designing for average users toward adaptive, personalized experiences, replacing click-based navigation with gesture-first patterns, and emphasizing data-informed decisions over intuition. Modern UX designers must understand business metrics and ROI to remain strategically relevant.
86
13
16
Article
Codemotion·35w
The world of programming has changed and with it the way to become a programmer
Programming has evolved from learning a single language to guarantee employment in the 1980s to navigating an overwhelming landscape of technologies, frameworks, and career paths today. The abundance of choices creates decision paralysis and impostor syndrome among developers, while AI tools like ChatGPT and GitHub Copilot are changing how code is written. Despite technological advances, the core skills remain curiosity, logical problem-solving, and tenacity. Modern programmers must learn to manage information overload and focus on fundamental problem-solving abilities rather than chasing every new technology.
83
14
17
Article
rryyqn·36w
Open Source Template for AI Support Chatbot
A free, open-source AI chatbot template built with Next.js 15 that includes rate limiting, bot protection, and customizable UI. The template uses Google's Gemini models and can be deployed quickly for customer support, knowledge bases, or lead capture. It features security through Arcjet, flexible configuration options, and is designed for easy deployment on Vercel with MIT licensing.
75
5
18
Article
ByteByteGo·34w
How Anthropic Built a Multi-Agent Research System
Anthropic built a multi-agent research system using an orchestrator-worker pattern with a Lead Researcher agent coordinating specialized subagents and a Citation Agent for accuracy. The system outperformed single-agent setups by 90% through parallel processing and dynamic adaptation, though it consumes 15x more tokens. Key engineering principles include proper prompt design, delegation strategies, effort scaling, and parallelization. Production challenges involve managing stateful agents, debugging non-deterministic behavior, and handling deployments without breaking running tasks.
74
19
Video
bigboxSWE·35w
DEVELOPERS ARE EXHAUSTED
Developer burnout is widespread, with 30% reporting job dissatisfaction despite good pay and remote work. Key factors include AI fatigue (despite 84% usage), extreme work hours like the 996 model, difficult job market conditions, technology stack overload, and poor cloud platform user experiences. While developers still love coding (70% code outside work), they're exhausted by constantly learning repackaged concepts rather than genuinely new skills.
72
11
20
Article
Stack Overflow Blog·35w
AI vs Gen Z: How AI has changed the career pathway for junior developers
AI has dramatically disrupted the traditional career pathway for junior developers and Gen Z professionals. Entry-level tech hiring decreased 25% in 2024, with 70% of hiring managers believing AI can replace intern work. Computer science graduates now face unemployment rates of 6-7%, higher than liberal arts majors. The traditional promise of coding as a secure career path has diminished as AI tools automate many junior-level tasks. However, there's potential for adaptation as the industry evolves to create new roles that complement AI rather than compete with it.
70
7
21
Article
Jeff Geerling·34w
I regret building this $3000 Pi AI cluster
A detailed analysis of building a $3000 Raspberry Pi cluster using 10 Compute Blades with CM5 modules, comparing its performance against more expensive alternatives. The cluster achieved 325 Gflops in HPC benchmarks but struggled with AI workloads, delivering only 0.28 tokens per second for large language models. Despite being energy efficient and compact, the Pi cluster offers poor price-to-performance ratio compared to traditional desktop hardware, making it suitable only for niche use cases requiring high node density or physical separation.
67
4
22
Article
Engineering Enablement·34w
How 18 companies measure AI’s impact in engineering
A comprehensive report examining how 18 major tech companies including GitHub, Google, and Dropbox measure AI's impact on software engineering. The study reveals specific metrics used to evaluate AI tool effectiveness, usage patterns, and productivity gains, providing practical guidance for organizations looking to assess their AI investments in engineering workflows.
67
1
23
Article
Javarevisited·37w
Why Software Architecture Skills Are the Key to Job Security in the Age of AI? (with Resources)
Software architecture skills provide job security as AI automates coding tasks. While AI can replace basic programming, architectural thinking—designing complex systems, making trade-offs, and guiding technical decisions—remains irreplaceable. The article promotes the Certified Professional for Software Architecture (CPSA) training as a structured path for senior engineers to transition into architect roles, highlighting significant salary increases and better career prospects for those with architecture expertise.
67
1
24
Article
Hacker News·35w
The Last Days Of Social Media
Social media platforms are increasingly flooded with AI-generated spam, bot accounts, and synthetic content that drowns out genuine human interaction. Engagement rates are plummeting as users grow exhausted with endless feeds of meaningless content optimized for clicks rather than connection. The traditional social media model is fragmenting as people migrate to smaller, private communities like Discord servers, group chats, and federated platforms. The future points toward decentralized, publicly governed digital spaces with transparent algorithms, user choice in content curation, and design patterns that prioritize intention over compulsive scrolling.
66
9
25
Article
Reinier·35w
CodeBuff: ClaudeCode KILLER! New AI Coding Agent is Quite Powerful, FREE, & Opensource!
CodeBuff is a new open-source, multi-agent AI coding assistant that offers both CLI and SDK integration for development workflows. It positions itself as a powerful alternative to existing AI coding tools like ClaudeCode, providing free access to developers.
65
4

See all AI archives