Vercel Hacked, Github Struggling! and why SWE may not be a long-term career anymore

Vercel Hacked! - Vercel disclosed a security incident involving unauthorized access to parts of its internal systems. The breach originated from a compromised third-party AI tool, allowing attackers to access a limited set of customer environment variables. Read the full incident.

Github struggling to keep up! - GitHub is tightening access to Copilot as new “agentic” AI workflows place unexpected pressure on its infrastructure. These autonomous coding agents consume significantly more compute, forcing the company to pause some sign-ups and introduce stricter usage limits.

Software Engineering May No Longer Be a Lifetime Career

AI is changing not just how we code but how long coding stays relevant

For years, software engineering has been seen as one of the safest long-term careers high-paying, intellectually engaging, and constantly in demand. The formula was simple: write code, get better over time, and your value keeps increasing. But with the rise of AI, that assumption is starting to look shaky.

The concern isn’t that AI makes developers less intelligent. There’s little evidence to support that. The real issue is subtler: when AI performs a significant portion of your work, you naturally learn less from doing that work. Debugging, writing logic, and solving complex problems are no longer fully manual processes. AI steps in, fills the gaps, and often delivers solutions faster than humans can reason through them.

In the short term, this is a massive productivity boost. Engineers can ship faster, fix bugs quicker, and build more ambitious systems. But over time, something important might be lost deep, hard-earned expertise. If developers rely too heavily on AI, their core technical instincts may weaken simply because they aren’t exercised as often.

Some argue this has happened before. Moving from assembly to high-level languages abstracted complexity, yet developers became more productive overall. That’s true but this shift feels different. AI doesn’t just abstract syntax; it abstracts thinking. It can suggest architecture, write entire functions, and even explain its own reasoning. This moves engineers from “builders” to “supervisors” of code.

And that shift has consequences.

In a competitive market, speed often matters more than mastery. Even if manual coding leads to deeper understanding, companies will likely prioritize engineers who can deliver faster using AI. This creates a dilemma: refuse AI and risk being outperformed, or embrace it and potentially sacrifice long-term skill development.

A useful analogy is construction work. Lifting heavy objects may harm the body over time, but it’s still required to do the job. Similarly, using AI might have downsides, but if it significantly improves output, it becomes unavoidable. The job evolves, whether individuals like it or not.

This is where the idea of software engineering as a “lifetime career” starts to break down. The profession may begin to resemble fields like professional sports highly rewarding, but with a shorter peak window. Skills can become outdated faster, and staying relevant may require constant reinvention rather than steady accumulation.

That doesn’t mean the field is dying. It means it’s changing. The most valuable skills may shift away from writing code toward evaluating it, designing systems, and managing AI-driven workflows. Engineers who adapt to this shift will continue to thrive.

The real risk isn’t AI itself it’s assuming that the old rules still apply.

Read full blog.

Google Cloud gets and upgrade - Google Cloud is pushing a major shift from traditional data systems to what it calls an “Agentic Data Cloud,” where data doesn’t just inform decisions but actively drives actions. Instead of static dashboards, this architecture turns enterprise data into a real-time reasoning engine that autonomous AI agents can use to plan, decide, and execute tasks.

Claude Gets a Memory Upgrade (Finally): Anthropic has introduced built-in memory for its Claude Managed Agents, enabling AI systems to learn continuously across sessions instead of starting from scratch each time. The memory layer stores knowledge as structured, portable files that agents can reuse, share, and manage with full developer control.

Buzz of the Week:

Ephemeral-to-Persistent Context Bridging

Ephemeral-to-Persistent Context Bridging is a technique that converts short-lived AI context (what the model “remembers” during a single session) into long-term, reusable knowledge. Normally, AI context disappears once a task ends, but this approach selectively captures important signals like decisions, patterns, or user preferences and stores them for future use. It acts like a filter, deciding what is worth remembering and what should be discarded. This makes agents more consistent and reduces repeated errors across sessions. It’s especially powerful in agentic systems where workflows span multiple steps and timeframes. In simple terms, it’s how AI stops thinking only in the moment and starts building continuity over time.

Things that launched. Things that went viral. Things you'll pretend to try.

nitric

nitric lets you define APIs, queues, and storage in code without worrying about cloud provider specifics. Deploy anywhere AWS, GCP, etc.

rspack

rspack is a blazing-fast alternative to Webpack written in Rust. Still under the radar but gaining traction for large frontend apps.

turso

turso is a distributed SQLite-based database designed for edge apps. Lightweight, fast, and surprisingly scalable.

Build Braincells, Not Just Features

This weekend’s read: tokenmaxxing as a weird new trend.

This week’s watch: Every Web Browser Explained in 10 minutes.

Meanwhile…