Alejandro Villamarin

Introducing Claude Sonnet 4.5 – the best coding model in the world and the strongest model for building complex agents. Sonnet 4.5 leads on SWE-bench Verified and can maintain focus for more than 30 hours on complex, multi-step tasks. It represents a major leap forward in computer use and agentic coding. Beyond the capabilities, this is our most aligned model yet, with substantial improvements in how the model communicates and behaves. We're also releasing the Claude Agent SDK—the same infrastructure that powers Claude Code—so developers can build agents with the same foundation we use. Sonnet 4.5 is available now as a drop-in replacement for Sonnet 4 at the same price. Full details at https://lnkd.in/gh_X5WjV

1235

51 comentarios

Since announcing my "High-Performance Node.js" workshop last week, I've received dozens of messages from enterprise developers facing the same critical challenge: "How do we identify event loop blocking in production before it becomes a crisis?" This is precisely the real-world problem separating junior developers from senior engineers. Let me share what I tell enterprise teams. The Three Warning Signs I Look For: 🚨 Response Time Spikes with Normal CPU Usage When your APIs start timing out but your servers aren't maxed out, you deal with event loop blocking. The server has capacity but can't process new requests because the event loop is stuck. 📈 Memory Growth Over Time Blocked event loops often cause promise queues to build up, creating memory leaks. Because of this single issue, I've debugged Node.js applications consuming 8GB+ RAM. Your monitoring should track memory trends, not just point-in-time usage. ⚡ Inconsistent Response Patterns Some requests complete in 50ms, while identical operations take 5+ seconds. This inconsistency is a classic indicator that something is intermittently blocking the event loop. The Real Solution Goes Beyond "Just Use Async/Await" Most developers think proper async/await usage solves everything. However, enterprise applications need comprehensive instrumentation, strategic worker thread implementation, and a deep understanding of Node.js internals. We'll tackle this in my workshop on September 29th in London. We'll debug real scenarios, implement proper monitoring, and build applications that perform consistently under production load. For enterprise teams dealing with Node.js performance challenges, this workshop bridges the gap between theoretical knowledge and production reality. 📅 September 29th | London | 4 Hours Intensive 🎯 Limited seats available Register: https://lnkd.in/d7A6wr2K What's your team's biggest Node.js performance challenge? I'd love to hear about it in the comments.

71

1 comentario

Accidentally discovered a new way to work with GitHub Copilot today: write new tests based off test coverage reports. Copilot (or I) run this command to generate a report of all the uncovered lines in the current diff: pytest --cov --cov-report=xml && \ diff-cover coverage.xml --html-report coverage_report.html && \ open coverage_report.html Then Copilot reviews the report, summarizes the uncovered situations, and proposes new test cases to add, or if the code is truly unreachable, suggests deleting it. Just another reason why you should add diff-cover (or similar tool) to your projects!

115

4 comentarios

“This is just a bad, terrible, no-good excuse for a protocol... It’s lacking in security, it’s needlessly complex, it’s not performant.” Criticisms of MCP? Actually these were all criticisms of REST when it was first proposed. Protocol bashing is a time-honored tradition in software. Every protocol sucks, except for the hypothetical ones that the critics haven’t gotten around to writing. REST had many detractors, and yet 80+% of all APIs are REST-based. MCP seems like a great bet to enjoy similar dominance in the realm of agentic software. All the criticisms of REST were valid, as are the criticisms of MCP. Very few protocols are perfect out of the box, but they’re good enough and software engineers are smart enough to fill in the blanks. MCP is still nascent. yes it has lot’s of imperfections, but it’s getting better every day. When MCP was published, servers had to be hosted locally, now they can hosted remotely. There was no framework for security, now there are guidelines around using OAuth 2.1 For a protocol, MCP has developed remarkably fast. Give it a couple years and the naysayers will be daily users. Bet on it #speakeasy #MCP

45

9 comentarios

Claude Sonnet 4.5 dropped 2 weeks back and proclaimed itself the best coding model yet. So we ran it through our internal agentic coding benchmark with real-world engineering tasks and codebases. Result: Yes - Claude 4.5 wins (but Codex is very close, and costs less than half) But something more interesting surprised me: About half the tasks each model failed were passed by the other. In other words, they're different types of coders. We deep dived one example: Sonnet 4.5: the craftsman perfectionist. Slow to err, obsessed with correctness, maybe a little neurotic about spacing but ultimately reliable. GPT-5-Codex: the hacker-engineer. Exploratory, error-prone, and a bit too eager to improvise. Full benchmark + example deep dive -> https://lnkd.in/emXSMiyd

101

9 comentarios