The Pulse #126: Startup asks for a lot, offers little → struggles to hire
Also: Cloufflare’s self-critical, transparent postmortem, additional signs of Big Tech becoming more cutthroat, and more.
The Pulse is a series covering insights, patterns, and trends within Big Tech and startups. Notice an interesting event or trend? Send me a message.
Today, we cover:
Industry pulse. Claude Code quietly becoming the best AI coding agent, Microsoft cutting AI data center investment, EA releases Red Alert code, developers don’t want GPUs, DeepSeek keeps open sourcing innovation, and a US DOGE software engineer forgets to make their GitHub account for work private.
Ads startup demands a lot & offers little → struggles to hire. An adtech startup using AI to generate ads seeks a founding engineer who will devote their heart and soul – and most of their awake time – to the company, and is struggling to hire. They’re now offering a $50K (!!) referral bonus. Feels like a classic case of overlooking that hiring is a two-way street: most standout devs don’t want to work 14 hours a day, 7 days a week.
Cloudflare’s self-critical, transparent postmortem. The leading content delivery network recently had its object storage go down for an hour, and only a day later published a detailed incident review, with the CEO personally taking responsibility for fixing the systemic issues which triggered the downtime. Clouflare is a refreshing exception for promptly sharing incident summaries, and showing it’s serious about continuously improving reliability.
More signs of “cutthroat” Big Tech. Google hints at a 60-hour workweek for workers in its AI group, and Meta fires 20 people for leaking internal information.
1. Industry Pulse
Claude Code quietly becoming the best AI coding agent?
Anthropic lowkey launched Claude Code, an agentic coding tool, recently. Feedback I’ve seen from engineers is that the tool is pretty good, with capabilities on par with other AI agents – including Devin (which costs $500/month).
Until now, Anthropic has built the best LLM for coding (Sonnet 3.5 and Sonnet 3.7). In hindsight, it was only a matter of time until they expanded by building an agentic coding tool that’s at least equal to what other startups offer. After all, most startups will build their agents on top of Sonnet 3.5 or 3.7, as well.
A criticism is that Claude Code is expensive, which should not be that surprising because AI agents are much more compute-intensive to run than LLMs. I’ll reserve judgement, Anthropic could turn into a major dev tools company if they keep building AI coding tools like this.