I owe Mozilla a thank-you. Really, I do. Maybe an Edible Arrangement? People like those. Some lil pineapples cut into stars on sticks and chocolate strawberries might brighten their day. For the note, I'm thinking something like:
Thank you for proving me exactly right.
XOXO MT
Eight months ago, in the fallout of Mozilla's fumbling of a Privacy Policy update, I wrote:
Mozilla is pursuing its primary objective, which is the survival of Mozilla. Its mission statement is more than broad enough to accommodate that, and Firefox is not a real priority. The community should accept that and stop waiting for Mozilla to be the hero they deserve.
Regrettably, I was unable to take my own advice on the last part. So here we are yet again, marveling at Mozilla's dedication toward eroding decades of good will in the community they purportedly serve. To quote one of my sacred texts, it's a focus and intensity normally seen only in successes.
Back in the present, we have Mozilla doubling tripling nthing down on this direction. First, with their announcement of "AI Window," a new feature (used very loosely) coming to Firefox which seems to emulate the user experience offered by AI browsers like Perplexity's Comet or OpenAI's Atlas. In other words, instead of performing search from the address bar and interacting with websites like browsers have done since they were invented, your first interaction will be with a language model prompt, which then mediates your experience of the web.
Not to gloat, but I told you so.
The response from the Firefox community has not just been overwhelmingly negative, it is universally negative as far as I can tell. At least among users willing to post on Mozilla's forums about the issue, which is absolutely a biased sample set. I have received some comments separately in support of Firefox, but they are countable and the vast, vast minority. Mozilla's core audience hates this move. At the very least, they would want all the AI components of Firefox to be opt-in, a choice that Firefox has been unwilling to make so far, instead enabling these new features by default.
What does Mozilla do? Temper the plan? Ease up on the forced features?
Nah, they do what any good corporate PR person would tell you to do when facing public backlash: post through it.
This post is a summary of Mozilla's new Strategic Plan, which is viewable in full here. I read it through a few times, and my brain nearly ripped in half from the cognitive dissonance involved. But I think it's worth examining Mozilla's claims carefully. They are:
- AI (by which they mean generative AI) is a transformative technology that will fundamentally alter how we interact with machines and the web.
- The current landscape is dangerous and controlled by big tech and "closed source" models.
- Mozilla should therefore pivot to develop and support "open source" AI implementations the same way they advocated for open web standards.
The strategy details the "what" and "how" of Mozilla's transformation in this direction. We're going to touch on some of those points, but let's begin with these big claims, affording Mozilla maximum benefit of the doubt.
Is generative AI a transformative technology? The Corpos sure seem to want it to be, although its actual usage seems mostly to be chatbot-related. All other attempts to use this trick in other realms have failed rather miserably. Microsoft, for example, wants you to talk to your computer instead of using a mouse and keyboard like a dinosaur. The results, unfortunately, are much worse than the Jurassic version of computer interaction. The pattern holds true across the board. Google's AI Overview continues to be an inferior provider of information than solid web search results. Also, as it turns out, people learn less from LLM output.
Even the AI browsers Mozilla wants to emulate have significant issues, vulnerable to old web vulnerabilities and new attacks against the models themselves.
Generative AI is transforming something, but I don't think it's the web, and I don't think it's for the better.
Which means their second claim is definitely true! The current landscape is dangerous, as I've been decrying for years.
But because Mozilla is convinced that generative AI is a force for good (with no evidence to back that claim up), they conclude that their mission must be to create "open source" alternatives to commercial ("big tech") offerings.
If you truly believe generative AI is a net good but with potential for significant harm, there are arguments to be made for ethical implementations. This same instinct is what propelled researchers from OpenAI to split and found Anthropic.
This, however, is an article of faith. It cannot be argued rationally because no empirical evidence exists to support it. The entirety of the belief is predicated on future potential—and it always will be, right up until the harms are so inescapably clear that even the most ardent of believers suffer because of them. Even then, not all of the Flock will lose faith. And as we now see, Mozilla leadership are not just the Flock, but Disciples.
Mozilla has had a conversion experience, while its core audience has not. This results in a schism of purpose.
The Actual Strategy
Digging into the plan itself, Mozilla's ambitions are remarkable. Mozilla has, as the plan notes, always measured itself against a "double bottom line" of mission and market success. However, it seems these two criteria are now separately defined by: "a. AI that advances the Manifesto; and b. diversifying revenue away from search."
Their specific goals include:
- All Mozilla orgs have a flagship AI product by 2028
- 10% year-over-year community growth
- 20% year-over-year growth in non-search revenue
- 3 Mozilla orgs have more than $25M in revenue (currently: 1)
- 10% year-over-year investment portfolio returns
I am dumbfounded by these goals. I can't even be snarky about them. They seem so disconnected from reality that I can't imagine how a Board arrived at them.
Let's go through them one at a time.
"Flagship" AI Products
What "orgs" are we talking about here? Historically there have been two Mozilla organizations: the Mozilla Foundation, which is the not-for-profit to which you donate to preserve the open web); and the Mozilla Corporation, which develops Firefox, makes deals with search engines, and creates other revenue-generating projects like, uh, Pocket.
There now exist three other for-profit subsidiaries of the Foundation, although these are not mentioned in the official listings. The first is MZLA Technologies Corporation, which is responsible for Thunderbird. I can't tell what else they do, if anything.
Another is Mozilla.ai, which has a much clearer purpose. This company produces AI products and services. So of these organizations, 3 of them need to have flagship AI products by 2028. What could that possibly look like?
There is also Mozilla Ventures, which is literally just a venture capital firm throwing money at AI projects that align with Mozilla's Manifesto.
Mozilla.ai has the easiest and clearest road, as their Agent Platformis already in early access and is absolutely a commercial product. MZLA has...Thunderbird? So AI-powered Thunderbird? That's what I can figure, although Thunderbird is not exactly known as an AI platform. Their new paid Thunderbird Pro service currently makes no mention of AI integration. That's a headscratcher. Mozilla VenturesAnd finally, we have the Corporation and Firefox.
Let's be as clear as we can possibly be. Mozilla is an AI company, and Firefox will be a flagship AI product according to this strategic plan. This is the focus for Mozilla, which means users of Firefox will only get more and more AI shoved down their throat, and likely fewer ways to avoid it.
Community Growth
I don't have access to Mozilla's "community" numbers, however they choose to define them. But if the reaction to recent changes is any indication, expecting growth of any kind isn't just optimistic, it's delusional. They are betraying the principles of their core use base in favor of their new god. That behavior is usually not rewarded by users or customers.
Financials
While I don't have access to community numbers, the Mozilla Foundation must disclose its financials, so that I can review—and you can too, if you're broken like me. I don't want to think about how much time I've spent reading Mozilla's 990s.
Looking at their revenue change from 2022 to 2023, we see a drop of 3% in royalties (search deals) and almost 15% in subscriptions and advertising. Let's also note that by these counts, royalties account for about 76% of Mozilla's annual revenue. That's by my own calculation on these disclosure, but Mozilla themselves cite 85% as share of revenue from search alone.
2022's numbers show a similar drop (down 3% from 2021) in royalties, but a 25% (!!) jump in subscriptions and ad revenue. My guess is ads, since I don't think the VPN service is raking it in, nor do I think a bunch of people suddenly signed up for Pocket before it died. This would certainly explain why last year, Mozilla went hard on their "privacy-honoring" advertising acquisition.
Zooming back out: the current business model is not delivering growth. So the pivot to AI is a bet that investment is out there for alternative sources of AI technology. It's also, tacitly, a bet that the AI bubble will last long enough to get competing products off the ground and attract investment before it's too late.
I...wouldn't be so sure.
As far as their investment portfolio's performance, Mozilla has changed their strategy significantly in the last 3 years, resulting in significant increases in dividend and realized gains in investments. What's the change in strategy?
At the end of 2022, Mozilla changed our strategy for managing our financial reserves. In prior years we took a purely defensive approach, investing solely in highly liquid fixed-income securities. Our revised approach is focused on delivering a total return to Mozilla after inflation, while maintaining sufficient liquid reserves to weather economic pressures and seize growth opportunities.
Translation: we invested more in stocks and less in bonds, T-notes, and CDs. What specifically they've invested in is unclear, but you can probably guess it rhymes with Blavidia. I'm sure they have a diversified portfolio, but if you believe (as I do) that the market is heading for a massive correction, this goal is unattainable and a dangerous target for a strategic plan.
Lastly, on revenue. Mozilla plainly has to diversify away from search, because search is dying. Google itself is trying to kill it, in favor of AI. If this succeeds, Firefox's primary revenue stream is drying up, and they know it. More than anything else, this is the reason for the pivot. As I've said before, the will to survive takes precedence over principle when choosing a path forward for any organization, even a not-for-profit with a stated mission.
These goals are not so much reasonable expectations as existential mandates. Either Mozilla approaches these targets in 3 years, or they may be staring death in the face.
Flawed Hypotheses
This strategy hinges on three stated hypotheses from Mozilla:
- A generational shift in human computer interaction is widening the gap between Mozilla’s products and trustworthy, user-centered experiences.
- A vibrant, successful and decentralized open source AI ecosystem is essential if we want independent tech players to thrive — and for innovation to come from everywhere.
- The growing need for sovereign, public interest AI which will only be met by governments and public interest tech players pooling resources and banding together.
That's...okay. We'll take it from the top.
Would you say generative AI is a "generational shift in human computer interaction?" The Corpos want it to be, but so far this hasn't taken place. Declaring it thus is the wish becoming the father of the thought. Maybe someday a functional language model will govern our interaction with computing machines, but that is nowhere near the case now, and the fundamental flaws in the technology preclude it from being so in the foreseeable future.
Would you say generative AI is "trustworthy" or "user-centered?" The people who implicitly trust generative AI are suffering from psychosis. It's a pathology. The model creators themselves tell you not to trust them! What are we doing here?
Y'know what is trustworthy? A goddamned URL bar that takes me to the website I want to go to. A search engine that shows me sources, ideally curated for quality.
Okay so hypothesis 1 doesn't pass the sniff test. On to number 2.
Can someone please explain to me what the hell "open source AI" is? Mozilla's helpful Strategy Wiki lists Mozilla's own products and investments under this category. Remember that for LLMs, you have two major "source" components: the dataset on which a model was trained, and the resulting vectors/weights file that comprises the model. Among them is HuggingFace, which is probably best understood as GitHub for AI. HuggingFace hosts both models and datasets used in ML/AI applications. It's about as close to open source AI as I can imagine.
Some of those HuggingFace datasets are really useful. Like, for example, the OCRed version of the Epstein Files. That's rad as hell, and I'm glad there's a place to share those things. It's even a goldmine for researchers like me, since datasets containing model jailbreaking prompts are available.
But let's be clear about Mozilla's value proposition of the "transformative" generative AI. These are not small models we're talking about; these are large language models that were trained on massive corpora of text. Those corpora are the "source." We know that the training data for frontier models comes from copyrighted material and material scraped without consent. We also know that code generated from scraped sources may well violate the licenses of that source code in reproduction.
In other words, there can never be an open source large language model when the sources are themselves violate of content usage agreements. For all the talk about "ethical" AI, Mozilla fails to address this original sin of the technology.
I'm sorry, I should say "nearly fails." In the "Threats" section of their SWOT analysis of their own strategy, they identify "Open models disappear" as a threat:
Big tech / China stop releasing open models. No public open source frontier models emerge. Mozilla’s strategy is obsolete / outflanked.
Okay so by "open source AI" you actually mean Qwen/Deepseek/Llama. Cool. Cool cool cool. These are open weight only, so the premise of open data goes out the window. And this threat gives away the fact that Mozilla can only succeed on the backs of frontier models. There is no real plan to "democratize" LLMs, nor can there be for the scale required.
This entire exercise is a farce. Yet again, Mozilla pursues a parasitical relationship with the corpos. It worked last time, right??
Hypothesis 3: the growing need for "sovereign" AI. We've already established that there is no large language model possible without corpo scale and investment, except perhaps with government support. So is that what Mozilla wants? State-sponsored LLMs? This hypothesis points in that direction, with Mozilla as the "public interest tech player" catching a percentage somewhere in the middle. Being a government intermediary is also probably not a safe position for anyone at this juncture, much less a tech company.
But also, what "need" are we talking about here? Why is there a need for any of this at all?
Again we encounter the fundamental schism of purpose between Mozilla trying to survive, and the mission its core audience believes in. You could imagine a Mozilla that decided, "Actually, the web was better without this dreck in it, and the experience of the web is not improved by moving users closer to it." You could imagine an organization that doubled down on true privacy, and a human-centered web. We'll never know what kind of funding streams such an organization could build, because Mozilla has chosen the machines over people. They have chosen quick revenue over long-term sustainability.
It's finally time you and I take the advice I offered before: let Mozilla die. It no longer serves its stated purpose.