Open Source Is Dead. Long Live Open Source.

Two supply chain attacks hit popular open source tools this week. Same week, a free Bloomberg Terminal clone shot to the top of GitHub Trending. Same week, OpenAI started talking about making open source "less of a security bargain," which is a polite way of saying "we think your free software is a liability and we'd like to sell you the alternative." Same week, a European coalition called Euro-Office started rallying open source as the weapon to kill European dependency on Microsoft.

Four things. One week. All contradicting each other. Welcome to August 2026.

I run a tech company out of Jönköping. We build AI agents, we ship software, we live inside open source code every single day. So let me tell you what I actually think, because most of what you're reading about this right now is either cheerleading or fear-mongering, and neither one is going to help you ship anything.

The Two Panics Are Both Real, and That's the Point

Panic one: open source got poisoned. The Register reported two separate supply chain attacks against popular open source tools this week alone. Not zero-days in obscure packages nobody uses. Popular tools. The stuff sitting in your dependency tree right now, three layers deep, that you have never once opened to read.

Panic two: open source is winning. Big time. Look at what's trending on GitHub right now. n8n sitting near 200,000 stars doing workflow automation that used to require a Zapier subscription and a prayer. opencode proving that a fully open coding agent can compete with anything venture-funded and closed. A free Bloomberg Terminal clone topping the trending charts, which should terrify every enterprise SaaS company charging six figures a year for a UI wrapped around market data. People are not just tolerating open alternatives anymore. They are actively hunting for them, and building them, at a pace that makes the "AI will kill open source" narrative from 2024 look almost quaint.

Both of these things are true at the same time. The mistake everyone is making is treating this as a paradox that needs resolving. It doesn't. Open source was never one thing. It's not a movement, it's not a company, it's not a philosophy with a board of directors. It's millions of independent decisions made by millions of maintainers, some of whom are funded by Microsoft and Google, some of whom are one unpaid guy in Belarus keeping a critical build tool alive between shifts at his real job. Of course it's simultaneously getting stronger and getting compromised. It's not a system. It's a population.

Why OpenAI Wants You Scared of Your Own Dependencies

I want to be direct about something. OpenAI positioning itself as "the adult in the room" on open source security is not a security initiative. It's a sales pitch. Digital Trends covered it straight, but read between the lines: a company that sells closed, hosted, managed AI infrastructure is telling you that open, self-hosted, community-maintained infrastructure is a security bargain you can't afford. That's not wrong, necessarily. It's just not neutral. Follow the incentive, not the announcement.

Here's the uncomfortable truth sitting underneath all of it: they might still be right, even while being self-interested. The attacks The Register reported this week are real. Supply chain compromise is the single most under-defended surface in modern software, full stop. But the fix isn't "buy closed" versus "use open." The fix is "know what you're running," which almost nobody actually does.

The Euro-Office Angle Nobody in Stockholm Is Talking About Enough

Meanwhile you've got Euro-Office rallying European companies to use open source specifically to escape Microsoft dependency. This is the geopolitical layer that most CTOs skip past because it sounds like a policy story, not an engineering story. It isn't. If you're running a Swedish company that has spent the last three years quietly migrating off US cloud tooling because of data sovereignty concerns, and now the tool you migrated to gets compromised in a supply chain attack, you haven't solved your dependency problem. You've swapped a corporate dependency for a maintainer dependency, and the maintainer doesn't have Microsoft's incident response team.

Sovereignty and security are not the same axis. Sweden keeps conflating them. So does the EU.

The Nordic Reality Check

Here's what it looks like from Jönköping instead of San Francisco or Brussels.

Swedish engineering culture is disproportionately open source native. We grew up on it. Spotify, Klarna, and half the fintech scene here have contributed back to open source for two decades, not because of ideology but because it was cheaper and better than the alternative. That instinct is good. It's also exactly why the supply chain risk lands harder here than people realize. A Swedish startup with three engineers and a Series A is running the same dependency tree depth as a company with a 40-person security team, but with none of the review capacity. Nobody at a 12-person shop in Jönköping is auditing a transitive dependency six layers down. Nobody at a 12-person shop anywhere is doing that.

Now compare that to the regulatory picture. The EU has the Cyber Resilience Act coming into force, which technically covers software supply chain requirements, and NIS2 is already live and applies to a lot more Swedish companies than founders think it does. But enforcement capacity is nowhere close to matching the ambition of the legislation. I've talked to founders who assume they're exempt from NIS2 because they're small. Read the actual scope. A lot of you aren't exempt. You're just unaudited, which is a different thing from compliant.

Compare that to the US, where there's no federal equivalent forcing the conversation at all, and companies are left to whatever their enterprise customers demand in procurement contracts. And compare it to what's happening in parts of Asia, where state-backed alternatives to Western open source infrastructure are being built specifically to avoid this exact problem, which is its own can of geopolitical worms I don't have space for today.

Sweden's actual advantage here isn't regulation. It's trust culture. We still mostly trust our institutions, our banks, our government digital services. That trust is an asset you can build security on, but it's also exactly the kind of asset that a well-executed supply chain attack exploits, because trust is what makes people stop checking.

Where This Actually Goes: 2026 to 2030

Here's my honest read on the trajectory, not the hype version.

The volume of AI-generated code entering open source repositories is climbing fast, and a huge percentage of it is being written by coding agents with no human review of the dependency choices those agents make. LangChain calling itself "the agent engineering platform" now instead of just a framework tells you where the center of gravity has moved. We are heading toward a world where an AI agent picks a package, another AI agent reviews the pull request, and the human in the loop is checking a dashboard once a week. That is an enormous attack surface increase disguised as a productivity win. As agent-to-agent software development becomes normal, supply chain attacks stop being a "clever attacker found a maintainer's stolen credentials" problem and start being a "an agent hallucinated a package name that a malicious actor had already registered and published with a payload" problem. This is already happening in isolated cases. It will not stay isolated.

On the AGI trajectory specifically: the closer we get to genuinely autonomous coding agents shipping to production with minimal human review, the more the entire concept of "trusted open source maintainer" needs rethinking. Trust used to be a person's reputation built over a decade of commits. In an AGI-adjacent world, reputation can be synthesized, forked, and impersonated at scale. The provenance question, proving that code actually came from who it claims to have come from, is going to be one of the defining infrastructure problems of the next five years. Whoever solves cryptographic provenance for open source packages at scale is building the next Let's Encrypt. Somebody in this Nordic ecosystem should be building that, not just talking about it.

The regulatory gap is going to widen before it narrows. The EU writes good policy slowly. Attackers move fast and don't file impact assessments. That mismatch is structural, not temporary.

What to Look At

If you want to actually do something about this instead of just having an opinion about it, here's where I'd start looking this week:

• n8n for workflow automation. If you're self-hosting anyway for sovereignty reasons, at least you control the attack surface directly instead of trusting a SaaS vendor's supply chain on your behalf.
• opencode as a case study in what an open coding agent looks like when the whole point is transparency over your development pipeline, not a black box you feed prompts into.
• system-prompts-and-models-of-ai-tools, which is a genuinely useful open repository showing you the actual system prompts behind commercial AI tools. Read it before you trust any closed agent tool with your codebase.
• Software Bill of Materials tooling. If your CI/CD pipeline isn't generating an SBOM on every build in 2026, that's not a nice-to-have anymore, that's the baseline. Start there before you start anywhere else.

What to Actually Do About It

Forget the ideology debate. Do these things:

Pin your dependencies. Not "mostly pinned." Actually pinned, with lockfiles committed and reviewed like code, because they are code. Run automated dependency scanning on every build, not once a quarter when someone remembers. Separate your sovereignty decisions from your security decisions, they are not the same conversation and treating them as one is how Swedish companies end up compliant with data residency rules while still running a compromised build pipeline. And if you're building anything with agents writing code semi-autonomously, put a human gate on every new dependency they want to introduce. Not on every commit. On every new dependency. That's the actual risk surface.

This is exactly the kind of paranoia we build into every MVP and fullstack build we ship at HEIMLANDR. Not because we're precious about it. Because the cost of fixing a supply chain compromise after launch is a rounding error compared to the cost of explaining to a client why their production database got touched by a package they never audited.

The Real Skill in 2026

The industry wants you to pick a side. Open source good, open source bad. Trust the community, trust the vendor. That binary is a trap built for people who write headlines, not for people who ship software. The actual skill right now, the one that separates founders who scale from founders who get burned, is holding both truths at once. Open source is winning against Big Tech's closed ecosystem, and that win is real and important and worth building on. Open source is also more compromised right now than it has been in years, and that risk is real and immediate and sitting in your dependency tree today. Neither cancels the other out. Building the paranoia into your pipeline, permanently, as a default rather than a reaction, is the only position that survives contact with reality.

Fredrik Brunnberg is the CEO of HEIMLANDR.IO, building AI and software solutions from Jönköping, Sweden. This is the daily HEIMLANDR briefing. If you found this valuable, share it with someone who builds things.