Devin
PulseAugur coverage of Devin — every cluster mentioning Devin across labs, papers, and developer communities, ranked by signal.
- developed by Cognition AI 100%
- developed by Scott Wu 90%
- employs Scott Wu 90%
- instance of Scott Wu 90%
- founded Scott Wu 90%
- founded Cognition AI 90%
- used by Mercedes-Benz 70%
- instance of Cognition AI 70%
- affiliated with Cognition AI 60%
- competes with Cognition AI 60%
- competes with Vercel 60%
- affiliated with Vercel 50%
- 2026-07-09 product_launch Cognition has launched SWE-1.7 for its Devin software engineering agent. source
4 day(s) with sentiment data
-
Vibe coding transforms software development with AI app builders and code editors · 2 sources tracked
Vibe coding, a new approach to software development that emerged in early 2025, has rapidly evolved into a significant industry. AI app builders allow users to describe products in natural language, generating functiona…
-
Andrej Karpathy's LLM Wiki concept challenges traditional RAG systems · 1 source tracked
A new concept called "LLM Wiki," popularized by Andrej Karpathy, proposes a shift from traditional Retrieval-Augmented Generation (RAG) systems. Unlike RAG, which processes and retrieves information on demand for each q…
-
Silicon Valley AI Talent Shifts: Versatility Over Specialization, Managers Decline
A senior AI engineer in Silicon Valley observes a significant shift in the tech industry, noting the decline of traditional middle management roles and a move away from the initial hype around AI tokens. He highlights t…
-
Cognition acquires Poke to enhance Devin's AI personality
Cognition has acquired Poke, an AI company focused on natural language interaction. This move aims to enhance Devin, Cognition's coding agent, by giving it a more conversational and engaging personality. The acquisition…
-
OpenAI launches GPT-5.6 models with costly 'Ultra' multi-agent mode
OpenAI has released GPT-5.6 Sol, Terra, and Luna models, along with the "Ultra" mode for Codex. This new mode, available since July 9, 2026, allows the GPT-5.6 Sol model to proactively break down tasks and delegate them…
-
Modal integrates Cognition's AI engineer Devin into its sandboxed environments
Modal has launched an integration called modal-devin, allowing the AI software engineer Devin, developed by Cognition, to run within Modal's sandboxed environments. This new capability, known as Devin Outposts, enables …
-
Cognition releases SWE-1.7 for Devin AI coding agent
Cognition has released SWE-1.7, an updated version of its Devin software engineering agent. This new iteration boasts near-frontier coding scores, indicating significant advancements in its ability to perform complex co…
-
Cognition CEO: Focus AI ROI on output, not token usage
Cognition CEO Scott Wu has criticized the trend of "tokenmaxxing," where companies incentivize employees based on AI token usage rather than actual output. Wu argues that this approach has led to excessive AI spending w…
-
Vercel CEO: AI agents need model separation for production
Vercel CEO Guillermo Rauch discussed the evolving landscape of AI agents, highlighting two primary use cases: coding agents and internal corporate agents for productivity. He emphasized the need to separate AI models fr…
-
Coding agent rules gain traction, highlighting AI development challenges
A markdown file containing four behavioral rules for coding agents has garnered significant attention, reaching 91,000 GitHub stars. This widespread interest highlights the current challenges and failures in the develop…
-
Cognition opens Tokyo office, targets Japan's legacy code market with Devin
Cognition, the creator of the AI coding assistant Devin, has opened an office in Tokyo, Japan. The company believes Japan's unique market conditions, including a shrinking engineering workforce and a significant backlog…
-
AI coding agents shift from prompt engineering to autonomous loops · 1 source tracked
The era of meticulously crafting AI prompts for coding tasks is fading, replaced by agentic workflows where AI agents autonomously execute a plan-edit-test-fix loop. These agents can manage tasks like migrating code, up…
-
AI coding assistants diverge in 2026 with Claude Code leading benchmarks
In 2026, the AI coding assistant market is defined by four key tools: Claude Code, Cursor, GitHub Copilot, and Windsurf. These tools have evolved beyond simple autocomplete, with significant advancements in context wind…
-
Ponytail AI tool cuts code generation by 54% with efficiency checklist
Ponytail is a new prompt-engineering tool designed to make AI code generation more efficient by encouraging agents to write less code. It functions as a plugin or skill that guides AI agents through a seven-step checkli…
-
AI software engineer Devin launches in Japan to transform development
Cognition AI is expanding its AI software engineer, Devin, into Japan. This move aims to revolutionize software development by leveraging AI. Devin is designed to handle complex coding tasks, potentially streamlining wo…
-
LLM prompt injection vulnerability rates vary widely across models
A security researcher tested five large language models (LLMs) for prompt injection vulnerabilities, finding that leak rates varied significantly from 0% to 90% depending on the model used. The tests revealed that disgu…
-
Devin CLI Extensibility Unlocks Advanced Customization for Developers
The Devin CLI offers advanced customization options beyond its basic REPL functionality. Developers can leverage extensibility features such as hooks, skills, subagents, and MCP servers to tailor the agent's behavior fo…
-
VS Code extension simplifies AI assistant tool integration
A new VS Code extension called "1 Click MCP Installer" by VePrompts simplifies the setup of Model Context Protocol (MCP) servers. MCP is an open protocol from Anthropic that allows AI assistants to connect to external t…
-
AI coding agents' pull requests show mixed success, high rejection rates
Two new research papers analyze the effectiveness of AI agents in software development, specifically focusing on pull requests. The first paper, "Toward Instructions-as-Code," found that while instruction files can guid…
-
Sarah Guo critiques AI benchmarks, open models, and silent model degradation
Sarah Guo's recent essay highlights key shifts in the AI landscape, questioning the future of open models and contrasting "model labs" with "agent labs." The piece also critiques the utility of current benchmarks, sugge…