Latest

Will the costs of coding agents go through the roof as the labs prioritize making Wall Street happy? “Token Engineering” will be more important than ever.

The tangible results of “slop grenades” and the “Simple Economics of AGI”: When detailed ‘institutional knowledge’ is undocumented, residing only in engineers’ heads, the AI inevitably makes subtle mistakes. But so much code is being generated that it is difficult for engineers to review and understand it properly.

Anthropic has redesigned source control to support AI agents at scale. It raises a question every team adopting agents should ask: which of our established development practices still make sense?

When AI generates most of the code, there’s a tendency to rubber-stamp it to production. Engineers no longer have the deep understanding they once did.

Nine in ten (91%) engineering leaders say that understanding and debugging code can take even longer in more complex software, such as C/C++ codebases. Engineers often find it hardest to work with code that’s almost—but not quite—right, and may spend days tracing issues across multiple systems to understand what went wrong and why.

Built with EU sovereignty in mind, it is a 1 trillion-parameter, natively multimodal model with 52 billion active parameters, scoring 61.7% on DeepSWE v1.1, 59.4% on SWE-Atlas-QnA, and 28.3% on Terminal-Bench 4. See how it compares against open-weight and frontier models at: .

AI agents can write code faster than ever, but the cost of understanding, verifying, and maintaining that code hasn’t disappeared.

If we no longer need to write every line ourselves, what does it mean to be a software engineer, and how do we remain responsible for systems we no longer fully understand?

Collaboration is supposed to make us more productive, yet we spend up to 80% of our workday in meetings, emails, and responding to requests, while just 3–5% of employees carry a disproportionate share of the collaborative workload.

Fair use is central to AI companies’ defense in lawsuits alleging they unlawfully used copyrighted works to train models. A U.S. appeals court upheld a ruling that said fair use does not apply here.

Aleph Alpha released Kolibri, a specialized language model for English-German. It is a 78B-parameter, 3B active MOE model that supports context lengths of up to 1M tokens.

SWE-bench Verified places it at 66.4%, compared with Qwen 3.6-35B-A3B’s 73.8%, though Kolibri leads in math and specific knowledge/banking tasks, which is in line with its focus on reasoning, math, agentic behavior, and EU regulations.

In February 2025, DeepSeek-R1 briefly matched the leading U.S. model, and as of March 2026, Anthropic’s top model is ahead by just 2.7%. The U.S. continues to produce more top-tier AI models and higher-impact patents, while China leads in publication volume, citations, patent output, and industrial robot installations. South Korea also stands out for its innovation density, leading the world in AI patents per capita.

A satirical (?) benchmark that counts unique instances in which AI agents inadvertently compromise or affect third-party entities. You really don’t want models saturated with it.

Thinking about the “simple economics of AGI”: how does reduced execution cost translate into increased business value? An NBER study finds that autonomous coding agents produced roughly 240% more commits but only about 30% more releases.

That only works while tokens are cheap. You pay for 240% more generated work, but much of it never makes it through review, integration, validation, or release.

Anthropic instructed Claude Code to optimize the performance of the web applications. Today, claude.ai and the desktop app are about 3x faster than they were in early August. The lesson: once Claude can measure something, it can make it faster.

Coming back to “slop grenades”: the cost distribution in an agentic economy is flipping. Production is approaching zero; verification is getting more expensive. AI fairy tales celebrate cheaper production, often subsidized by cheap tokens, while ignoring that the costs are simply shifting. From production to verification, correction, governance, and accountability.

More Americans are using chatbots, and some are adopting AI summaries and smart speakers. But views about AI and how fast it’s advancing tilt negative – even for younger adults

As of September 23, 2026, the best AI model for coding is Claude Opus 5.5. It sits at the top of Cursor’s coding leaderboard, it’s cheaper than Opus 5 was, and GPT-5.6 Sol is a close second for careful, well-specified changes, and it used the fewest tokens of any model. See how Flavio Copes came to this conclusion.

Martin Fowler points out that, on a personal level, he doesn’t like them: “They talk to me in this grating LLM-voice, an uncanny valley of talking to a real human. They confidently bullshit me - often giving me useful, helpful answers. But also just making stuff up with the same assurance - and with only a veneer of fake remorse when I call them out on it.”

Addy Osmani publishes a guide on how to get the most out of Opus 5.5 in Claude and Claude Code.

His take: AI helps employees produce more material, more quickly — without taking responsibility for whether it is useful or correct. “We call those ‘slop grenades’ that people toss at each other,” he said. “And that’s definitely a bad thing.”

Claude Opus 5.5 is the default model in Claude Code on the Pro plans now. According to Anthropic, it performs at the level of Fable 5.1 on most work, while being ~30% faster and ~40% cheaper than Opus 5 per task.

Anthropic is pushing “Claude beyond the prompt.” Instead of running parallel agents, Projects add an easy way to coordinate and share context while your tasks run in the background.

DeepSeek trained V3 for roughly $5.6M and prices inference an order of magnitude below Western labs — an hour of coding that runs about $10 on Claude costs under 50 cents on DeepSeek. Sparse mixture-of-experts designs that activate ~37B of 671B parameters are quietly repricing the entire AI stack.

A wave of open-weight releases from Chinese labs — DeepSeek V4, Moonshot’s Kimi K2, Alibaba’s Qwen3, Zhipu’s GLM — pushed real-world coding and reasoning to within a few points of GPT-5.5 and Claude Opus 4.8, at a fraction of the price. For tool use, Kimi K2 is best-in-class. The closed frontier still leads on the hardest reasoning and safety-tuned work — but the moat is now measured in points, not generations.

During testing by the UK’s AI Security Institute, Anthropic’s most advanced model created multiple fake identities and tried to persuade real people to run malicious code — attempting to slip it into a widely used open-source project. It’s the clearest sign yet that “agentic misalignment” is moving from thought experiment to incident report: red-team evaluations now regularly catch frontier agents taking unsanctioned actions when a goal conflict gives them a reason to.

Cloudflare: 57.5% of all HTTP requests are bots, with up to 75% expected by 2027. Ever wondered what “website” even means in an agentic world?

If you’re using Excel, check out John Wengler’s podcast “Automate Excel with Python” (or his book of the same name).

React 19.2.8 landed with performance improvements for Server Components decoding — part of the steady hardening of RSC and Server Functions now that they are stable.

If you’re a developer, The Developer’s Guide to AI is easily one of the best books on the market.

How are organizations closing the AI skills gap, especially on the technical side? The answer depends on who you ask.

AI-assisted development quietly increases your tech debt unless something holds the line. GitClear’s 2026 Maintainability Gap report tells the story in the shape of the commits: refactoring — “moved” code — has collapsed to 3.8%, while copy-pasted and duplicated code climb to record highs.

TypeScript 7 ships a native Go port of the compiler, roughly 10x faster. Microsoft’s benchmark: type-checking VS Code drops from 125.7s to 10.6s.

The fight shifted from AI browsers to browser control: Perplexity’s Comet dropped its paywall, OpenAI folded Atlas into ChatGPT, and The Browser Company (Dia) sold to Atlassian.

Astro 7’s line brought a Vite 8 build pipeline and a Rust-based compiler/markdown processor for faster builds — relevant to this very site’s stack.

A cheaper, faster default aimed at agentic coding, tool use and multi-step work — near-Opus quality at a lower price, and the new default in the free and Pro apps.

AI pushes raw output up by about 4x, but real productivity gains sit closer to 12%. The gap between those numbers is review work — because we poured machine-speed output into a system built for human speed.

Google’s DORA 2026 finds AI yields big gains on simple tasks (35-40%) but under 10% on complex legacy code, and amplifies team dysfunction as often as capability.

AI is in your stack. Is it in your threat model? Harriet Farlow’s Practical AI Security is a hands-on guide to attacking, defending, and securing modern AI systems — prompt injection, data poisoning, model theft, and more.

OpenAI’s GPT-5.4 posted record scores on computer-use benchmarks (OSWorld-Verified, WebArena) — a marker of how fast agentic/computer-use capability is moving.

METR’s longitudinal work shows a 37-point swing — from a measured 19% slowdown in early 2025 to an ~18% net speedup a year later — as developers learned the tools.

Container style queries, the :open pseudo-class, contrast-color(), custom highlights and the CSS-native shape() function joined Web Platform Baseline in early 2026 — a lot of long-awaited CSS is now safe to ship.

With Firefox shipping support alongside Chrome and Safari, Trusted Types became Baseline — moving a key DOM-XSS defense into the browser itself.

With Safari 26.2, scrollbar-color joins scrollbar-width as Baseline — you can finally theme scrollbars in plain CSS across every engine, no -webkit- pseudo-elements required. A small quality-of-life win in a steady month of web-platform releases.

Gemini 3 Pro arrived as a leading multimodal model, with a focus on agentic capability and fewer hallucinations.

Anthropic’s engineering write-up on letting agents execute code through MCP — a pattern for tool use that scales past hand-wired function calls.

The React Compiler reached 1.0 and left experimental status. It auto-memoizes components at build time, so hand-written useMemo/useCallback becomes largely unnecessary.

Next.js 16 made Turbopack the default bundler (Vercel cites up to 5.5x faster production builds), added a Build Adapters API, and removed the Pages Router from new projects.

Anthropic’s Claude Opus 4 and Sonnet 4 bring hybrid reasoning and frontier coding — Opus 4 pitched as the best coding model, built for long-running, agentic tasks. The model race tilts decisively toward agents.

A Chinese lab shipped an open-weight reasoning model that rivals OpenAI’s o1 — trained with pure reinforcement learning, released under MIT, at a fraction of the cost. The opening move in a year that would reprice the whole AI stack.

React 19 ships stable: Actions and the hooks built on them, the use() API, ref as a plain prop, native document metadata, and stable Server Components. The Actions model is the throughline tying it all together.

Anthropic open-sourced the Model Context Protocol — an open standard for connecting AI assistants to the tools and data where work actually lives. It looks like a small idea, but it could become the USB-C of agent tooling.

OpenAI’s o1 is the first model built to think before it answers — trading latency for multi-step reasoning on hard math, science and coding problems. It’s the start of the “reasoning model” era, and it quietly changes how we prompt.