Same quality, a quarter of the cost: Should DeepSeek Flash be your model of choice?

DeepSeek Flash offers comparable quality to pricier models at a fraction of the cost, making it a cost-effective choice for running agentic tasks at scale.

Rob Willoughby, Simon Maple · 8 min read · 9 Jun 2026

The hidden cost of agentic software development: why context engineering matters

AI token costs in software development are rising, impacting budgets. Context engineering is crucial for managing these expenses and ensuring efficient resource use.

Paul Sawers · 7 min read · 9 Jun 2026

Anthropic details how attackers are weaponising Claude Code — but says AI will ultimately give defenders the edge

Anthropic's report reveals how attackers weaponize Claude Code for autonomous cyber attacks, but suggests AI advancements will eventually favor defenders.

Paul Sawers · 8 min read · 8 Jun 2026

AI Native DevCon Day 2: From Agent Demos to Operating Models

AI Native DevCon Day 2 explored operating models for AI-native delivery, focusing on context pipelines, agent behavior metrics, and organizational ownership.

Rohan Sharma · 12 min read · 3 Jun 2026

The model's solved, now comes the hard part: Reviewability as the bottleneck

AI engineering shifts focus from model development to ensuring system reviewability, emphasizing manageable task sizes for reliable and governable outputs.

Paul Sawers · 9 min read · 2 Jun 2026

AI Native DevCon Day 1: Making AI Agents Ready for Enterprise

AI Native DevCon Day 1 focused on making AI agents enterprise-ready, emphasizing reliability, skills as code, and adapting platforms for agent integration.

Rohan Sharma · 12 min read · 2 Jun 2026

AI Coding Agent Accuracy: Opus 4.7 vs 4.8

Opus 4.8 matches Opus 4.7 in accuracy but improves efficiency, solving tasks in fewer turns and at lower costs, highlighting differences beyond headline metrics.

Rob Willoughby · 9 min read · 29 May 2026

Opus 4.8 tops the LLM leaderboard with 95% on skill evals

Opus 4.8 leads the LLM leaderboard with a 95% skill evaluation score, surpassing Opus 4.7 and Composer 2.5 Fast, despite being the slowest model tested.

Simon Maple · 8 min read · 29 May 2026

Why We're Changing Our Default Eval Model

The default eval model is changing from Claude Sonnet 4.6 to GLM 5.1 to reduce costs without losing signal quality, focusing on skill evaluation over model specificity.

Rob Willoughby · 9 min read · 29 May 2026

We ran Composer 2.5 and 2.5 Fast across 11 skills. Surprisingly, Fast won.

Composer 2.5 Fast outperformed Composer 2.5 across 11 skills, scoring higher and running 32% quicker, while costing the same, challenging typical speed-quality trade-offs.

Simon Maple · 6 min read · 28 May 2026

Don't Make Your Agent Guess

The article discusses the importance of tool design in agent systems, emphasizing that prompts alone are insufficient for ensuring agent safety and reliability.

Matthias Lübken · 11 min read · 27 May 2026

The Reinvention Of The Dev Team

Explore how AI is reshaping dev teams, challenging traditional roles, and introducing new dynamics in software development, focusing on speed, safety, and value.

Hannah Foxwell · 11 min read · 26 May 2026