#25 - AI coding tools: Blessing or curse? 🤖

Hey readers! 👋 The AI coding landscape is evolving at breakneck speed, with new tools launching weekly and adoption rates soaring. This week we're diving into Google's revealing DORA report showing nearly universal AI adoption among developers, alongside some surprising insights about who's actually benefiting from these tools. Plus, exciting new releases from OpenAI, GitHub, and more that are reshaping how we write code!

🔍 This Week's Highlights

Exa-code aims to eliminate LLM code hallucination by indexing over 1 billion documentation pages, GitHub repos, and StackOverflow posts. Given a query, it performs a hybrid search and returns token-efficient chunks that significantly improve accuracy. In evaluations, exa-code outperforms all popular web search tools and is available for free on GitHub.

How AI is reshaping developer teams explores how AI tools are revolutionizing software development beyond just writing code. The conversation between Peter O'Connor and Ryan J. Salva highlights how AI is streamlining deployment, operations, and team collaboration, enabling smaller teams to operate effectively with reduced overhead. – Katja Skafar

OpenAI's ChatGPT Codex 2.0 is powered by the advanced GPD5 High model and available in three versions: cloud-based CH GBD Codex, lightweight Codex CLI, and IDE-integrated Codex Extension. Key features include detailed pull request reviews, token compression, modular coding, and step-by-step planning for complex workflows. – Julian Horsey

GPT-5-Codex is now available in Cline, bringing a 400K token context window with pricing matching GPT-5 ($1.25 per million input tokens, $10 per million output tokens). It features adaptive reasoning that adjusts effort based on task complexity and variable thinking that works perfectly with Cline's thinking slider. – @nickbaumann_

Google's 2025 DORA report reveals that 90% of software development teams now use AI daily, spending about two hours actively working with AI tools. While over 80% report productivity boosts, trust remains mixed—30% trust AI output little or not at all, while 24% trust it greatly. The report emphasizes that successful AI adoption depends on organizational adaptation, not just tools. – Darryl K. Taft

AI helps strong dev teams and hurts weak ones, according to Google's DORA report. AI acts as an amplifier, magnifying the strengths of high-performing organizations and the dysfunctions of struggling ones. Seven key practices drive AI's impact, including clear AI policies, healthy data ecosystems, strong version control, and quality internal platforms. – @ZDNET

🚀 New Tools & Features

Free stealth model "code-supernova" is now available in Cline, designed specifically for agentic coding tasks. It supports a large 200k token context window and multimodal inputs, including images like screenshots and diagrams, making it well-suited for iterative, tool-heavy coding workflows. – @nickbaumann_

Progress Software has expanded AI coding assistants across its Telerik and Kendo UI libraries, now supporting most major UI component libraries. These AI assistants integrate with popular AI coding tools like GitHub Copilot and Claude Code, providing context-aware code generation that helps developers increase productivity by up to 30%. – Progress Software Corporation

Warp Code aims to streamline the process of turning agent-generated code from prompts into production-ready software. It integrates a lightweight editor and inline code review directly into the Warp terminal environment, allowing developers to make quick edits without switching tools. – Paul Sawers

GitHub Copilot CLI is now in public preview, entering the terminal-based AI coding assistant market alongside tools like Claude Code and Codex CLI. It integrates with GitHub's Models backend, defaulting to Claude Sonnet 4 but allowing users to switch to GPT-5. – Simon Willison

Hugging Face has released an extension that integrates open-source large language models directly into GitHub Copilot Chat within VS Code. Developers can now test models like Kimi K2, DeepSeek V3.1, and GLM 4.5 without leaving their editor. – Tom Smith

zAI has introduced affordable GLM Coding Plans that integrate with Cline, offering developers access to the GLM-4.5 AI model for coding assistance at just $3/month for the Lite plan (120 prompts per 5-hour cycle) and $15/month for the Pro plan (600 prompts per 5-hour cycle). – @nickbaumann_

💡 Best Practices & Insights

Cline's guide to Cursor alternatives without usage limits introduces Cline as an open-source AI coding agent with strong VS Code integration and support for local, cloud, or hybrid AI models. It provides a decision framework for selecting AI coding tools based on privacy, usage intensity, and budget. – design

Code review with agents best practices include reviewing after each subtask to avoid review debt, prefixing clarifying questions with "Just a question" to avoid unintended edits, using #TODO(agent) markers for batch refactors, and verifying with git diff and tests before merge. – @augmentcode

GitHub's Spec Kit enhances AI-assisted software development by combining detailed specifications with human oversight. Unlike simpler AI coding tools, Spec Kit integrates software engineering best practices, guiding users from project specification to a working prototype with dependencies installed. – Simon Bisson

The Year of the Agent podcast explores how agentic engineering evolved from a fringe experiment to a movement reshaping software development. The discussion between @sjmaple and Reuven Cohen covers what makes "agentic engineering" a new profession, how hive minds mirror nature's most efficient systems, and why large companies are "AI-washing" agent tech. – @ainativedev

📊 Industry Trends & Research

Bain & Company's research reveals that AI coding tools have delivered only unremarkable savings and productivity gains so far. While AI can perform up to 40% of a developer's tasks, coding constitutes less than 40% of a software engineer's workday, and adoption among developers remains low. – ITPro

Google executive Ryan J. Salva emphasizes that "there's no hiding from AI if you're a software engineer." With AI tools being used for code generation, review, documentation, and testing, software engineers are shifting their focus from raw coding to product design and problem-solving. – TOI Tech Desk

Engineering-centric AI is transforming enterprise software by making it more flexible, adaptive, and predictive. Engineers play a critical role in integrating AI safely, ensuring ethical use, security, and continuous improvement. – Express Computer

Hacker News discussion on Google's report reveals skepticism about actual productivity gains from AI coding tools. Many developers report AI-generated code often requires extensive review and debugging, and that AI can amplify organizational weaknesses. – working for the social media companies

🔧 UI & Workflow Improvements

Agent Tabs put Thread, Tasks, and Edits in the chat header, each with full vertical space for better readability and focus. This UI update is available in VS Code v0.549.1+ with IntelliJ support coming soon. – @augmentcode

CodeSpell AI offers a streamlined interface that consolidates all code recommendations in one place. Features include grouping recommendations by file, the ability to accept or reject changes individually, and eliminating the need to switch between tabs. – @codespellai

Phase 3 of incremental refactoring introduces cross-file impact analysis, parallel remote agents for bulk edits, and "Next Edit" workflows that automatically update dependency trees. – @augmentcode

CodeRabbit now integrates with Codex to allow autonomous AI-driven coding, reviewing, and issue fixing directly within development workflows without human intervention. – @coderabbitai

Made with ❤️ by Data Drift Press - Hit reply with questions, comments, or feedback!