AI coding agents
16 published articles
Coding agents · Devin's two-model harness
Paying 2x more per token made Devin's Fusion 9% cheaper
Devin's Fusion runs a frontier lead model beside a cheaper sidekick, each with its own context. Devin's benchmark tables show task costs dropping as much as 46%, and one case where a lead costing twice as much per token produced a 9% cheaper session.
2026-09-26
Developer tools: autonomous coding agents
Cursor wants to run your agents and host the code they write
Cursor shipped Projects and Origin, a code host that can be the source of truth for repositories created inside it. Both arrive in beta, and the gaps in the announcement matter as much as the features.
2026-09-26
Alibaba Cloud on securing AI agent logs
Your AI coding agent just logged a database root password
AI coding agents read .env files and tool output, then write what they find into audit logs. Alibaba Cloud's Simple Log Service offers three interception points, from on-device masking to envelope encryption with a split key.
2026-09-15
AI Coding Agents
Alibaba's Qoder bets coding agents will be won on the harness, not the model
Alibaba Cloud's Qoder Agent Desktop is an argument as much as a product: coding agents will be won on the harness, not the model. Tasks now replace files as the unit of work, and Alibaba points to internal numbers showing a 40% cut in agent input tokens. No independent verification has been published yet.
2026-09-15
AI coding security
Qoder moves code security from the release gate to the keystroke
Qoder Desktop and CLI now run security checks at three depths, from live keystroke screening to cross-file data-flow analysis before a merge. Alibaba Cloud reports roughly 60% better vulnerability detection and a 35% to 45% drop in security feedback during code review.
2026-08-13
AI Agents & DevOps: STAROps inside Qoder
A one-sentence question in Qoder ends the 40-minute root-cause hunt
Alibaba Cloud's STAROps plugin brings natural-language root-cause diagnosis into the Qoder IDE. Its demo takes a P95 spike from under 60ms to 1.9 seconds, returns an evidence chain at 80% confidence, and ends with an auto-created merge request. Traditional troubleshooting ran 40-plus minutes across five platforms.
2026-08-09
Agentic IDEs: betting on task boundaries
Alibaba says Qoder 1.0 cut agent input tokens 40%. Alibaba ran the test
Qoder 1.0 rearchitects the AI coding IDE around task boundaries, claiming a 40% cut in agent input tokens and a 22% drop in dissatisfaction on Alibaba's own three-day A/B test. No independent replication has been published yet.
2026-08-09
Meta AI
Muse Code: Meta's crash-resilient bet in the coding agent race
Meta's Muse Code brings a crash-surviving terminal coding agent to macOS and Linux, powered by Muse Spark 1.2, a model co-trained with its own harness for long-horizon tasks. Meta says larger, more capable models are already on the way.
2026-08-06
AI Development
Coding agents are leaving your local machine: Alibaba and Mistral go remote
Alibaba Cloud's Qoder introduces remote delegation for coding agents, letting developers offload long-running tasks to cloud sandboxes. The feature, paired with a new knowledge engine, pits Qoder directly against Mistral's remote agents in a race to decouple AI-assisted development from the local terminal.
2026-08-03
AI Agents
Nobody Can Say What Their AI Coding Agents Actually Did Today
Enterprises spending heavily on AI coding agents can't see what those agents actually did, the benchmarks measuring their improvement may be overfit to public test sets, and the code they write isn't reviewed by default. Three separate problems with the same root cause: adoption outran the tooling to observe it.
2026-07-30
Agent Observability
Your AI coding agents are burning millions, and nobody knows where
Enterprises are spending heavily on AI coding agents with zero visibility into their actual behavior. Alibaba Cloud's open-source LoongSuite Pilot aims to fill that gap, but the core issue runs deeper: nobody can track what these agents are doing.
2026-07-26
WWDC 2025
Apple's walled garden just got a gate with an AI-shaped key
Xcode 27 brings agentic coding with Anthropic, Google, and OpenAI models, plus new intelligence frameworks that let apps use models from Apple, Claude, Gemini, and others. Core AI lets developers run full-scale LLMs on-device.
2026-07-25