All posts by rbowen

Community Over Code Europe 2026, Glasgow


Event: Community Over Code 2026
Dates: October 11–14, 2026
Location: Glasgow, UK
Website: https://communityovercode.apache.org/events/glasgow-2026/

Community Over Code is the Apache Software Foundation’s flagship conference. Glasgow 2026 ran October 11–14, with a three-day hackathon (Mon–Wed) in the Wee Dram Room alongside the main program. There were also two events (Lakehouse Day, Airflow Contributor Day) on the 10th, before the main event started.

Things that I was particularly interested in

Non-code participation

The strongest cross-cutting thread of the week: open source is more than code, and the projects that thrive treat non-code participation as first-class. Peter Zaitsev (Percona): “users give open source its purpose” — not every user can contribute code, but they can contribute their story, their frustrations, and their complaints.

Evan Rusackas (Superset) built an entire talk around ten distinctly human roles in the age of AI.

I gave at talk, with Kris Freedain, about non-code contributions – in particular, how project maintainers should very intentionally create specific opportunities for non-code work, and go out of their way to welcome and recognize the people that do it.

Governance maturity and leadership succession

Dawn Foster spoke twice (Leadership in OSS; Good Governance), and the themes recurred all week: define the path to leadership; document values, mission, and scope; have an explicit, non-punitive path to emeritus; define how problematic people are removed; and start simple, adding governance complexity only when it solves a real problem.

Documentation as the scalable on-ramp

Docs came up as the highest-leverage, most-neglected lever for onboarding. Highlights: Google’s “Docs Advisor” tool concept for writing better docs – at https://github.com/google/opendocs – provides templates and tools for common docs patterns

I was introduced to the concept of “friction logging”, which is where you journal what was hard, or didn’t work, as a means to capture that beginner’s mind and find places that could be made easier for those beginners. I’m going to start doing that!

AI’s double edge for open source integrity

A theme echoing the PyCon Africa data (70% of developers using AI daily): AI lowers the cost of producing code but not the cost of earning trust — “code is getting cheaper; trust is not.” Practical responses discussed: add a small good-faith hurdle before accepting contributions so you filter AI-gamed contributions from genuine ones, and lean on humans for judgment, mentoring, and the roles AI cannot fill.

Justin McLean noted AI is genuinely useful for anonymizing lessons without the bias of preconceived opinions about specific projects, people, or companies.

Keynotes

We had a number of really great keynotes. I want EVERYONE to watch Russell Spitzer’s keynote when it’s published.

Russell Spitzer — cross-project collaboration

Russell’s keynote was all about how Snowflake recognized the importance of meaningful community engagement in the projects that they were depending on. Since this is what I do in my day job, having someone else say it so eloquently was very affirming, and is something I want all of the managers I work with to watch. Really a great presentation.

Vlad Miricioiu

Transitive and “phantom” dependencies are largely invisible to CVE tooling, and with vendored code most dependencies are “phantom”. This was a fantastic overview of how deep understanding of your supply chain and dependencies is critical to security and sustainability, and how badly we are collectively doing with this. This is also worth watching if you care about those topics.

Stephen Walli

Stephen gave a talk about his last few years in teaching college students to do software engineering better, in collaboration with various open source projects. While I had heard the talk before (and that’s why we invited him!) it was great to see it with a new audience. His points about students self-mentoring, and that all software is social software, landed very well with the Apache audience.

Brian Baglow

Brian talked about the video game industry, which was almost all new information for me. I really don’t know how to summarize here. He talked about the engineering patterns in game development, and how they have evolved over the decades. And about the money that’s in that world, and how it’s distributed. And how it compares to open source development. And many other topics. Definitely worth watching, and hugely informative.

I’ll post links to all of these referenced talks when they’re published.

The Hackathon

It’s been several years since we had a formal Hackathon at Community Over Code, or Apachecon before that. We have almost always provided a *space* for people to collaborate on stuff, but we haven’t done any promotion beforehand for years.

This year, we worked with communities to bring a task list, and promote to their communities. This resulted in three days of a full room, with real work getting done, and new people being onboarded.

I hope to have more stories to tell about this over the coming weeks.

We had 20 projects participate formally – https://events.apache.org/events/2026/community-over-code/hackathon.html – as well as several projects drop by to use the space for more ad-hoc work.

PyCon Africa, Kampala, 2026

(Yes, it’s long. I have many many thoughts, and even left some of them out!)

Event PyCon Africa 2026
Dates October 7–9, 2026
Location Speke Resort Munyonyo, Kampala, Uganda
Website https://africa.pycon.org/2026/

PyCon Africa is the pan-African Python conference, hosted in 2026 at the Speke Resort Munyonyo on the shores of Lake Victoria in Kampala, Uganda. It draws Python developers from across the continent — students, early-career engineers, founders, and a growing professional developer community — alongside a smaller audience from outside of Africa. I attended on behalf of AWS to assess the event, build relationships in the East African developer community, and evaluate it as a candidate for ongoing AWS sponsorship and participation.

Why

According to the opening keynote, Africa now has 700,000+ professional developers, with Uganda growing at roughly 8% year over year and Kampala emerging as a regional tech hub. (Every presentation I attend about this has very different numbers. I think the main point is that it’s A Lot.) Uganda alone has an estimated 36.3 million smartphone users. This is a young, mobile-first, fast-growing builder population — and it is forming its cloud habits.

Meanwhile, to give credit where due, Google has been working hard here for 15+ years. Every developer in Africa knows about Google Summer of Code. (I don’t think that’s an exaggeration.)

There is also an AI inflection underway: roughly 70% of developers at the event reported using AI tools daily. That is both an opportunity and a risk — the risk being blind trust and the erosion of verification and debugging skills. Whoever helps this generation of developers build well with these tools earns durable mindshare.

And Python is the default language, both in AI work, and general application development work. (I’d love to hear stories about other communities with similar passion and excitement in Africa!) So PyCon seems to be the place to find the people who are creating Africa’s technology future.

PyCon now runs in 12 African countries with several more onboarding; Ariane Djeupang serves as the Africa chair. A recurring theme from community leaders, which Ariane said in her keynote, was “Go, so you can be in the room where decisions are made — but come back and lift those you left behind.” This is a community that is deliberately building its own leadership pipeline.

The story that a number of people told was that as soon as African developers become successful, they get a job in Europe or North America. The desire to build a local talent pool which stays to build the African economy, with companies owned and staffed by Africans, was repeated in a number of presentations.

What I Saw

It was interesting to compare audiences between West Africa and East Africa. Those of us from outside tend to treat Africa as a monolith, but it’s not. Africa is not a country. And culture varies considerably between West Africa, East Africa, Sahara, and Southern Africa, and even those categories are too large.

The folks at OSCAFest in Nigeria, last year, were at a party. They were excited about the technology, about the people on stage, and about making connections. But, also, in the sessions, they were deeply engaged, very professiona, and absorbing the technical details, and looking for practical tips to implement right away.

The audience in Togo seemed earlier career. They were excited about the learning opportunities. They were looking for places to get involved in Python, and other open source projects. And they were looking for a way to break into the larger technical career world.

And the audience in Uganda were clearly established professionals. They were masters of their crafts, looking for ways to up-level even further. But there was also a lot of frustration about the availability of jobs, particularly without leaving home. There were a lot of remote workers, benefiting people far away, rather than close to home, and that seemed a source of frustration for many of them.

These differences showed up in how people engaged with the talks, and in what kinds of talks held the audiences, vs the talks where folks were just scrolling on their phones. That’s worth remembering when proposing future talks – understand the local culture, and (particularly for foreigners) do your homework – don’t assume that Africa is all the same. It’s not.

Africa’s Talking

Mark Ssembajjwe‘s keynote on Friday was particularly great. Mark works at Africa’s Talking, a place for African startups to get a start. They operate in 23 countries with 300,000 developers on its platform — a mobile-first development platform with a Python SDK, mobile-payment infrastructure, and a marketplace for distributing applications. They run hackathons, workshops, and office hours. Their goal is to build African tech companies, owned and staffed by Africans, which provide jobs for Africans so that they don’t have to go abroad to find jobs.

This, and efforts like it, are an outgrowth of the frustration with decades of Africans working to enrich the lives of foreigners, and was very exciting. I wish them much success.

Local models and local data

Roland Gafana (AI Studios) presented “The Last Mile is Data” — a Google-funded (Yep, again, Google) effort building local-language models (Ganda Gemma, Swahili Gemma 1B, Luganda models), open-sourced and published on Hugging Face, plus OpenData.ug, which asks people to donate their data rather than having it extracted. Google funds local-language open models because it drives consumption of their services. It is a great example of strategic open source investment in an emerging market, and it’s another reflection of the desire for technological sovereignty – a topic I’ve talked about before on this blog.

Meeting with community leaders

Right before I left (KLM changed my flight and I had to leave before the event ended. Boo!) the event planners gathered several community leaders from across the continent, to meet with me and John Kimani to talk about why we sponsor events, what those events can do better for us, and what we (the cloud providers) can do better for these events and their attendees. They are, of course, looking for sponsorship for future events. But more than that they are investing in their local economies, and believe that young developers, getting technical skills and then NOT immediately leaving the continent, is the way to ensure their nations’ future.

This was such a great opportunity, and I feel greatly honored that they were willing to do this. The people in that little gathering are changing Africa, but they’re so humble and selfless about it.

This was the main reason that I wanted to go – to meet these exact people.

Why you should care

Africa is the future of open source. There are genuine hurdles to development in Africa – infrastructure and opportunities still lag behind much of the rest of the world – but the talent there is just staggering. And the desire to build something by Africans, for Africans, is a motivation much stronger than parts of the world that are primarily doing outsourced work for foreigners. This is definitely a part of the world to watch.

I put some thoughts in this video, but I always feel like I’m not adequately able to convey the real passion that I encounter in events in Africa.

If you have an opportunity to go to an open source event in Africa, grab it.

See also:

Driving in Kampala

Uganda National Mosque – a view from above the city

Comparing today’s Greenland treaty with the 1951 version

1951 Treaty vs. 2026 Agreement: What Each Side Gains and Loses

Disclaimer: This analysis was assisted by an AI assistant (Claude) and should be verified against the source treaty texts.

1951 treaty (Yale Avalon Project): Defense of Greenland: Agreement Between the United States and the Kingdom of Denmark, April 27, 1951

2026 agreement (Danish PM’s office): Agreement to Amend and Supplement the 1951 Defense Agreement, September 22, 2026 (PDF)

This new agreement, signed September 22, 2026 in New York, doesn’t replace the 1951 treaty — it amends and supplements it. The 1951 framework remains in force; the 2026 text layers new rights, obligations, and constraints on top. Here’s the balance sheet:


🇺🇸 What the United States Gains

Area Detail
Expanded basing rights The 1951 treaty allowed defense areas only “as the two Governments may from time to time agree.” The new deal explicitly authorizes the U.S. to modernize and expand Pituffik Space Base (formerly Thule), plus establish new bases at Narsarsuaq and Mestersvig — named locations, not just a vague future possibility.
Open-ended future bases Either party can now propose additional defense areas. If the Permanent Committee can’t agree within 90 days, the issue escalates to deputy-minister then minister level — a forcing mechanism the 1951 treaty lacked.
Unmanned installations Entirely new category: the U.S. can request unmanned military installations outside defense areas under an “accelerated governmental approval” process. The 1951 treaty had no such concept.
Submarine/undersea access The new Article VI(ii) explicitly grants U.S. public vessels undersea access to and movement within Greenland’s territorial waters “without restriction except as mutually agreed.” The 1951 treaty mentioned surface and air access but was silent on submarines.
Golden Dome missile defense The preamble specifically references allowing U.S. military access for “establishment of a Golden Dome defense system” — a named strategic program blessed in treaty text.
Investment screening veto Article X effectively gives the U.S. a veto over non-NATO/non-EU foreign investment in Greenland’s sensitive sectors (critical infrastructure, resource extraction). States or investors from non-NATO, non-NATO-partner, non-EU countries can’t have control, significant influence, or access to non-public info unless the Parties agree it’s safe. This locks out Chinese and Russian investment in Greenland’s rare-earth minerals.
Anti-espionage cooperation New Article VII requires Greenland/Denmark to “work together [with the U.S.] to counter espionage” and keep “close contact with relevant U.S. authorities” — a formalized intelligence-sharing obligation with no 1951 equivalent.
Security perimeter around bases The new Article VII gives the U.S. the right to raise concerns about construction or land-use changes near defense areas, and the parties must “mutually determine what steps may be necessary.” The 1951 treaty had no buffer-zone concept.
Permanence & independence lock-in Article XI says the agreement has no end date (the 1951 treaty lasted “for the duration of the North Atlantic Treaty”). More importantly, if Greenland becomes independent, it must (i) remain in or join NATO and (ii) assume all U.S. defense rights from this agreement. This survives independence.
Third-state military exclusion Article IX bans any non-NATO state from establishing military installations or maintaining a persistent military presence in Greenland. No such provision existed in 1951.

🇺🇸 What the United States Loses / Concedes

Area Detail
Greenlandic contracting preference Article IV(iv) requires the U.S. to award contracts for building, maintaining, and servicing defense areas to Greenlandic sources “to the maximum extent possible.” The 1951 treaty had no such requirement — the U.S. could use its own contractors freely. This means more expensive construction and slower timelines.
Respect for Greenlandic way of life Article VI now explicitly conditions all U.S. military access rights on “respecting the Greenlandic society and the Greenlandic way of life, including hunting, fishing, and other traditional, cultural, historical, and other future activities, to the maximum extent possible.” The 1951 treaty had a vaguer obligation to respect “laws, regulations and customs pertaining to the local population.”
No unilateral base creation While the U.S. gains a faster process, it still can’t establish new bases unilaterally — the 90-day escalation mechanism still requires mutual agreement, just with a deadline.
Sovereignty reaffirmation The U.S. formally reaffirms “the sovereignty and territorial integrity of the Kingdom of Denmark” — effectively taking Greenland acquisition off the table as a matter of treaty record.

🇩🇰🇬🇱 What Denmark/Greenland Gains

Area Detail
Greenland as a treaty party The 1951 treaty was between the U.S. and Denmark. Greenland wasn’t a party — it was talked about, not with. The new agreement is explicitly tripartite: Denmark together with the Government of Greenland. Greenland signs separately.
Self-determination recognized The preamble explicitly recognizes that “the People of Greenland constitute a people pursuant to international law with the right to self-determination” and references the Greenland Self-Government Law’s independence procedure. The 1951 treaty said nothing about Greenlandic self-determination.
Sovereignty reaffirmed The preamble reaffirms “the sovereignty and territorial integrity of the Kingdom of Denmark.” Given the political context (Trump’s interest in acquiring Greenland), getting this in writing is meaningful.
Economic benefits from bases The mandatory Greenlandic contracting preference (Article IV(iv)) is a direct economic gain — U.S. military spending on construction, maintenance, and services must flow to Greenlandic businesses first.
Environmental protections preserved The preamble reaffirms the 1991 MOU and 2004 Joint Declaration on environment. New base proposals must include “mitigating measures” accounting for these environmental commitments. The 1951 treaty was silent on environment.
Cultural/traditional rights protected Explicit protection for hunting, fishing, and traditional activities in the face of expanded military operations.
Stronger NATO commitment Article II supports “a stronger NATO engagement in the Arctic, including as regards planning, presence, exercises, and joint intelligence gathering.” Denmark gets formalized Arctic security cooperation, not just a U.S. bilateral arrangement.
Denmark’s own defense posture Article VIII acknowledges Denmark’s strengthened Arctic security posture (land, maritime, air, space) — codifying Denmark’s own military buildup as part of the framework.

🇩🇰🇬🇱 What Denmark/Greenland Loses / Concedes

Area Detail
De facto veto over foreign investment Article X effectively gives the U.S. a say over who invests in Greenland’s resources. Denmark/Greenland can no longer freely court Chinese investment in rare earths or critical infrastructure without U.S. agreement. This is a significant sovereignty concession.
Locked into NATO permanently If Greenland goes independent, it must join/remain in NATO and carry forward all U.S. defense obligations. An independent Greenland can’t choose neutrality or a different security alignment. The 1951 treaty simply lasted “for the duration of NATO” — it didn’t bind a future independent state.
More U.S. military footprint The 1951 treaty resulted in one major base (Thule). The new agreement green-lights at least two more named locations plus an open-ended expansion mechanism with accelerated processes. More boots, more radar, more missiles on Greenlandic soil.
Submarine access in territorial waters Granting undersea access is a meaningful concession — it means U.S. submarines can operate freely in Greenland’s territorial waters, something not previously codified.
Security buffer zones The U.S. can now object to construction or land-use changes near defense areas. This limits Greenlandic development options in the vicinity of bases.
Espionage cooperation obligation Denmark/Greenland must maintain “close contact” with U.S. intelligence on counter-espionage. This is a formalized obligation that may constrain diplomatic flexibility with countries the U.S. considers adversaries.
No end date The 1951 treaty was at least theoretically terminable when/if NATO dissolved. The new agreement has “no end date” and can only be changed by mutual consent. It’s effectively permanent.

The Big Picture

The U.S. got what it wanted strategically: expanded, essentially permanent military access to Greenland with new bases, missile defense infrastructure, submarine access, investment screening powers, and a guarantee that even an independent Greenland stays in the U.S. security orbit.

Denmark/Greenland got formal recognition of Greenlandic sovereignty and self-determination, Greenland’s status as a treaty party (not a colonial subject), economic benefits through contracting preferences, environmental and cultural protections, and — perhaps most importantly — the U.S. stopped talking about buying Greenland and put Danish sovereignty in writing. The sovereignty reaffirmation and self-determination language is the political price the U.S. paid to get its military expansion.

It’s essentially a grand bargain: military access and strategic lock-in for the U.S., in exchange for formal recognition of Greenlandic identity and economic participation, plus the implicit promise that acquisition is off the table.

Slave House

August 30, 2026
Wood House (Woold Homé)
Agbodrafo, Togo

(Read more about the Wood House here.)

Slave House

I hardly need to tell you,
an American
about slavery

But he did.

They stayed under the floor —
a space
the same height as the underdeck
where half of them would die
on their way to
their new life
in a far away place they could not imagine

The women getting some
temporary liberty
above the floor in the evenings

If you know what I mean

Then back below
for two, three,
maybe four weeks.

This was the master’s bed
his chair
his safe.

The German tourists
anxious to establish that
the rapes
happened before they came
to rape the country

Yes, it was the Scots.
then the Portuguese and the English

And, of course,
their countrymen

Because money is money
and people are just people.

My Togolese friends
did not know this
about their country.

Of course, I knew it about mine.

We crawled under the floor,
each squatted in the dark
thinking of the crimes of
our ancestors
much too recent to be
entirely forgiven

FOSSY 2026 — Notes from Vancouver

FOSSY (Free and Open Source Software Yearly) just wrapped its fourth year — this time at the University of British Columbia in Vancouver, BC. A few hundred developers, community managers, foundation staff, and advocates, people who care deeply about how free and open source software gets made and sustained.

I attended primarily for the Governance and Community Sustainability track. But as usual the hallway track was where I got the most value.

Here’s some of the talks that were particularly valuable.

Sustainability at the PHP Foundation

Elizabeth Barron from the PHP Foundation gave a talk that moved well beyond “we need funding” into specifics. The PHP Foundation (established 2021) now employs 13 contractors who produced 42% of all PHP commits in 2025. That was a pretty big surprise to me. At the ASF, we take the opposite approach – all code is by volunteers, but we staff infra and other operational things.

The PHP Foundation is at Open Collective, who, if you don’t know, are a fiscal hosting agency (ie, they just handle your money) and make both taking and spending money very easy for PHP.

One thing that she talked about that was particularly interesting to me was the idea of attempting to articulate how much it costs to run a project, and approaching sponsors with that number, so that they clearly understand where the money is going. This seems obvious in retrospect, but I don’t see many folks doing it.

She also mentioned thanks.dev which is a means to route funds down the dependency tree so that the less visible projects get some of the pie as well.

This felt like one of the more mature “here’s how it actually works” sustainability talks I’ve seen — less hand-waving, more spreadsheets.

The Art of the Fork

I gave a talk on planning for forks as a disaster recovery plan, and Edward Ly from NextCloud complemented it nicely with “Art of the Fork: Reclaiming Trust.” His framework for evaluating forks boiled down to three questions:

  1. What distinguishes my fork from the original?
  2. How hard is it to switch?
  3. How will the original project change in response?

Simple questions, but the kind you want written on a whiteboard before you commit engineering resources to either side.

Code of Conduct Research — With Actual Data

Shurui Zhou presented research on what happens after a project adopts a code of conduct — not opinions, but data from 100,000+ projects, comparing contributor patterns 3/6/9/12 months before and after adoption against matched control projects that did not adopt one.

Key findings:

  • 77.4% of CoCs were never updated after initial adoption
  • Both short and long term: increased new contributors
  • Short term: some disengagement of existing contributors (quantitative only — reasons not investigated)
  • Transparency reports (incidents, resolutions, response times) help signal accountability

The conversation in the room about why existing contributors leave short-term was interesting, if speculative. Based on anecdote, a number of people in the room suggested that the initial disengagement of existing contributors wasn’t necessarily a bad thing, but was people self-selecting to leave because they didn’t want to have to behave.

The data itself is valuable for anyone making the case to adopt or improve enforcement.

University OSPOs and the Student Pipeline

Stephanie Lieggi from UC Santa Cruz talked about lessons from running a university OSPO. Two things I noted:

  • CURIOSS.org — a consortium spanning 30+ university OSPOs internationally
  • UC Open Summit — an annual conference on a UC campus focused on academic open source

The persistent question: how do real projects get in front of students in a way that leads to sustained contribution?

AI + Free Software Development

A panel on generative AI in free software surfaced the tension well:

For: Help maintainers handle the overwhelming flood of inputs. Do the mundane so developers can focus on creativity. Lower the barrier for new contributors who historically were gated by coding expertise alone.

Against: Unearned confidence in wrong solutions. Massive walls of LLM-generated text are hard for humans to review. AI-generated PRs may be increasing maintainer burnout rather than reducing it. And while contributor numbers are at an all-time high, the number of maintainers isn’t budging — which raises the question of whether we’re creating contribution without creating ownership.

This lack of ownership is, IMHO, just an acceleration of the drive-by PR phenomenon that Github started, and must be addressed if we want longer term sustainability.

Encrypting your server

Romeo Solano gave a great talk on practical ways to encrypt your filesystem on a server that might not be physically accessible to you.

This was of particular interest to me because Romeo attended my “How to give a better presentation” talk last year at FOSSY, and we had talked some about ways to improve his stage presence. Romeo was funny, engaging, and informative, and his presentation was impossible to look away from. I was truly impressed.

I attend a LOT of conference presentations, and I don’t often add one to my list of “This is how presentations should be done.” This was one of them, for sure.

Overall

FOSSY is small and opinionated. That’s the feature. The conversations you have here — with foundation directors, long-time maintainers, university OSPO leads — happen because it’s 300 people, not 3,000. If you work on open source governance, sustainability, or community strategy, it’s worth your time.


FOSSY 2026 took place August 6–9 at UBC in Vancouver, BC. More at fossy.us.

“Trump’s Pathological Aversion To Truth”

This article was just fascinating to me.

https://www.huffpost.com/entry/john-kennedy-trump-anxious-unexpressed-thought_n_6a6f91e2e4b03026f4f00de1

“He grows anxious when he has an unexpressed thought,”

When my son was little he struggled with keeping his inner dialogue inner. If he thought something, he had to say it. If he didn’t say it out loud he would become very agitated. If we were at an event where he couldn’t talk continuously, like a school orchestra concert, for example, he would become more and more agitated until he had to scream.

We left a lot of events early.

And once spoken an idea had to be defended forever no matter how ludicrous. If he stated his firm belief in an absurd concept and you dared to criticize that idea, you were calling him a liar. And this was true even after he discovered proof that he was wrong. Once spoken, the idea became a hill to die on, a core belief to defend.

But then, as children do, he grew up.

In this article, John Kennedy describes the president of the United States of America in every way identical to my 7-year-old son. Unable to keep a thought inside and contemplate whether it was true. Unable to change his mind once presented with truth and evidence.

On the one hand, it helps me understand what’s going on with this strange, sad, pathetic man.

On the other hand, it makes me all the more contemptuous that at 80 he has not learned what my son did at 8.

UN Open Source Week 2026

I also recorded a video version of this post if you’d prefer to watch rather than read: https://youtu.be/VPIf6dUKwKA

Last week I was at the United Nations Headquarters in New York City for UN Open Source Week — a gathering that, even just by its name, tells you something about how far open source has come. This isn’t a developer conference. It’s not a vendor expo. It’s the United Nations convening governments, NGOs, academics, and technologists to talk about open source as a matter of global policy.

The event ran June 22–25 at UN HQ in the ECOSOC Chamber, then moved on June 26 to the IBM Building at 1 Madison Avenue for a community-led “Open Source Community Day” hosted by Red Hat. It’s not the kind of event I usually attend.

The thread that ran through almost every session was technological sovereignty — specifically around AI. Who controls AI infrastructure controls the future.

Alberto Gago, Director General of Spain’s Agency for the Supervision of AI (AESIA), said:

“Decisions affecting all of society are being passed to organizations controlled by a handful of private actors answerable to no electorate. Transparency is not only not guaranteed — it’s not even possible.”

That’s not a developer complaining about vendor lock-in. That’s a government AI regulator saying the architecture of modern AI makes democratic oversight structurally impossible.

One of the more interesting ideas I encountered was the collapse of the distinction between Sovereign AI (nations controlling their own AI infrastructure) and Portfolio AI (diversifying workloads between hyperscalers and local compute). These are the same question at different scales. A country asking “how do we avoid dependence on foreign tech companies?” is asking the same question as an individual asking “can I run some of this locally?” The answer in both cases is open source — open weights, open models, open tooling. It’s the only path to that kind of independence.

The highlight of the event was the “OSPOs for Good” session, opened by Omar Mohsine (UN Open Source Coordinator) and featuring government ministers speaking about open source as a development instrument.

Amal El Fallah Seghrouchni (Head of Government in charge of Digital Transition, Morocco) said:

“Open source is the instrument of sovereignty.”

Angellah Jasmine Kairuki (Minister of Communication and Information Technology, Tanzania) talked about a national determination to

“Change from being a passive consumer of technology to an active participant.”

These are government ministers framing open source as national policy — as the tool that prevents their nations from becoming permanent technology colonies of whoever builds the largest proprietary models. The Next Einstein Initiative (“Will the next Einstein come from Africa?”) kept surfacing as a motif — talent exists everywhere; infrastructure and access don’t. Open source is positioned as the equalizer.

Another topic of great interest to me was The RGAF – The Responsible Generative AI Framework (RGAF), developed by the LF AI & Data Foundation, which defines nine dimensions for responsible AI: Human-centered & Aligned, Accessible & Inclusive, Robust & Safe, Transparent & Explainable, Observable & Accountable, Private & Secure, Compliant & Controllable, Ethical & Fair, and Environmentally Sustainable. It aligns with the EU AI Act, NIST, Singapore, and China frameworks and is an attempt to create a shared vocabulary across all of these, specifically for the open source generative AI context.

More info on RGAF:  https://lfaidata.foundation/blog/2024/11/18/responsible-ai-pathways/

UN Open Source Week isn’t a technical conference. You’re not going to learn a new API. What you get is perspective — a window into how the rest of the world sees open source and what they need from it. When Tanzania’s Minister of Communication and Information Technology stands at a podium in the ECOSOC Chamber and says open source is how her country avoids becoming a permanent technology colony, that reframes every commit, every pull request, every licensing decision.

More photos on Flickr.

Dark Carnival

When figuring out my read order for Bradbury Summer, I discovered a book that not only have I not read, I’ve never even heard of.

Dark Carnival was Bradbury’s first book, in 1947.

It’s a collection of short stories, some of which I’ve read in other collections (Homecoming, and Uncle Einar, are in October Country) but most I have not

I assume there are other things of his that I’ve not read, and I’m really looking forward to discovering them!

Winesburg, Ohio

The preface to The Martian Chronicles says I should also read Winesburg, Ohio, by Sherwood Anderson. Books that recommend other books is perhaps my largest source of recommendations over the years.

The Illustrated Man

I just finished The Illustrated Man, which I haven’t read in decades. I was worried it wouldn’t stand up, but it was every bit as amazing as I remembered, but with the extra depth of reading it as an adult.

Particularly wonderful was “The Rocket” in which a broke junk dealer arranges for his kids to travel to Mars in a scrap metal rocket.

Bonus, about half of these stories ended up as Twilight Zone episodes!

Next, The Martian Chronicles, which I have read many times over the past 40 years but is always wonderful.