A lot of people in the comments say that the supposedly better examples in the article are still “horrible” because they can tell they’re AI and point to things that the AI did “wrong” and a human artist would do better.
That may be true if you’re working with a top of the line artist, but in my experience the average freelance designer online is significantly worse than AI. I’ve worked with dozens of budget friendly designers on Fiverr and consistently get way worse results than I’ve gotten from AI models. Will a great human designer produce better output than an AI? Maybe, but a small local event or business probably won’t have the budget to hire them or know where to find them. It’s cheaper, faster and easier to use AI to generate something that’s good enough for their purposes.
People also point out all the little flaws in the “good” examples and cite them as obvious tells that they were AI generated. I think it would be interesting to see if people could reliably distinguish between AI generated flyers and human made flyers. Sure, sometimes it’s obvious like when the AI adds those weird, cartoony people. But if the model is told not to create designs like that and stick to the styles used in the second half of the article, I expect people would have a much harder time distinguishing the human designs from the AI designs.
It’s a tale as old as time — people don’t understand that marketing and branding are just as important, if not more so, than the product. Jev is exceptionally-well branded. Anyone can look at the webpage and understand it, and the implications, instantly.
OPs “marketing” is a single post on Reddit titled “ Predicting sales conversion probability from conversations using pure Reinforcement Learning”. Can you understand what that means? I can’t, and I consider myself reasonably technical. Is it obvious it has the same implications as Jev? Again, no idea. And it was just a single post on a subreddit that I don’t even browse! I see people on this thread saying “Jev is just BERT”. Sure, and Dropbox is just a ftp account mounted with curlftpfs!
I do feel bad for the author for finding something cool and being unable to brand it. But the full definition of “product” INCLUDES being able to coherently communicate it. In some sense the branding is just as much the “breakthrough” as the model.
The biggest problem that I have with most AI posters is that they are too busy, you have to look for information instead of seeing it popping right away.
Amateurs posters of before AI didn't have this problem. White background, the important info in big letters, maybe one clipart or to. So I'm not talking about how pro artists could make a better job, but even the small local non-profit used to produce better posters. Not in the sense of "better looking", but in the sense that they were more readable.
At least the examples of this blog are more readable than what we usually see.
* Different brain regions have different functions: ancient knowledge.
* Different regions contain distinct cell types and molecular programs: also well established.
* Anterior and posterior brain structures may trace back to fundamentally separate progenitor lineages that were independently specified very early in evolution and development: that's the interesting new result.
The headline should have been: "Human forebrain and hindbrain arise from distinct embryonic progenitor lineages" - not that we have "two organs" in the brain.
I think the main gripe that people had with Jev and Typesafe was the language used when they launched. To me personally it seemed like a parody/con/shady at first.
"Breakthrough", "our research went in another direction" , "Two years in stealth", "System One thinking model", "Jev can't hallucinate", "RLCD","We are doing very cool stuff, but we will have to hire you to tell you", - these are some of the things that they said on their website on the launch blog.
I had used versions of bert to achieve the same functionality years ago. But to me it seems like they were able to trick the VCs with "can't hallucinate" etc.
To the above author, kudos for sharing your work and making it open. Something like this shouldn't be closed in the first place when it has been available for so many years
It’s not complicated - the default style signals low effort. Why should I get excited about the village fete if the organizers put in the barest minimum of effort?
But there’s another reason it drives people crazy - it’s low effort trying to present as high effort. If the sign was simple Comic Sans then it would at least be charmingly honest.
I find it kinda funny that even "smartest", most capable models often have a very hard time going beyond surface-level, top-of-mind associations, when faced with creative tasks.
Of course, a "Japanese Minimal Poster" has a sakura and a stylized flag of Japan, duh
A human designer would probably think that "Japan -> sakura" and "Japan -> Japanese flag" are way too obvious, too banal, too stereotypical, and, most importantly too boring. And then they would probably sit for a while and try to think of something fresher and less clichéd.
But AI has absolutely no problem with going with the cheesiest, most overused trope.
Hi! I'm the proprietor of onionfutures.com. I'm actually not a member of UChicago for Our Futures and came up with the idea independently, but we've since connected and they handle our Chicago deliveries.
I played around with Jev last night and did it for classification tasks that I used Gemini 2.5 flash lite with.
It’s a bit faster and bit cheaper, but this is compared to LLM. The consistency was nice to see, BUT, as someone who trained NLP models prior to LLMs, it’s just BERT with more data. I can see why people would want ready made one shot classifier, and I can see the value of sending multiple classifier in one call, but I wouldn’t call it breakthrough. And I believe many labs will replicate it in no time and might have it as part of their harness.
I see it as a wake up call for the tech community to go back to basics for most tasks instead of relying solely on generic LLMs.
From TFA: "Eric Schwitzgebel writes that . . . There’s a huge cognitive difference between nodding along while reading something and actually productively generating a text. Two reasons: First, once the text is on the page, it’s easy to passively let the approximate word suffice, rather than thinking about word choice in the same effortful, active way we do when generating prose de novo. Second, as I suggested above, I doubt that human beings, even experts, have a good sense of all the factors that shape word choice -- everything they’re being sensitive to. You would have phrased it slightly differently, and even if you don’t know that, or why, a different signal is sent and received."
This is the best articulation I've seen of why simply reviewing and copy-editing does not provide remotely the same value as writing from scratch. I spent a considerable amount of time over the past two weeks reviewing and improving a work document that was the output of an LLM. Given the number of people involved and the final level of effort, I'm firmly convinced that writing it manually would have been faster and resulted in a higher quality product. Getting the wording right matters.
The people behind this are a real (if silly) student organization at UChicago and Northwestern campaigning for the legalization of onion futures trading. Sometimes they will aggressively hand out free onions to passerby on campus. It's quite entertaining.
If you want people to pay you for your software, stop writing it for free. Conversely, if you write it for free, don't expect people to pay you for it. Otherwise you are no better than someone at an intersection with a bottle of Windex and a squeegee who, unsolicited, cleans a windshield and then demands the driver to pay for it.
The original authors of Free Software and open source were career academics and others who were paid to do other things, or were sponsored by scientific and defense research grants. I don't know how anyone got the nutty idea that you could make money on FOSS itself. Practically every time someone has tried to make money on FOSS it has failed.
(Edit: this comment previously ended with "...from Netscape on down.")
I’m reminded of that famous debate between Poincaré and Hilbert at the International Congress of Mathematicians in Paris in 1900. It was then that everyone decided to follow Hilbert’s path, and proof came to be valued more than intuition. I think modern math at school and at applied university kind of lost this intuitive part.
I try to teach my students that mathematics is, first and foremost, a very precise language of communication. It’s sometimes amusing to ask those who don’t like math to do without it entirely, just to see how much harder it becomes to describe the things around them.
Second thing I tell them, formulas are the essence of mechanisms in their purest form. And in this form, they’re much easier to grasp and mentally manipulate. It always amused me, after taking a mechanics course, to imagine that for any formula, you could visualize a mechanism or process that implements it.
And third thing, I suppose, the ability to verify one’s own statements as proof. Although, of course, mathematicians would probably tear me apart here for my heresy:sorry, I’m not a mathematician, but an engineer. You can make mistakes by using incorrect assumptions, but at some point, analysis itself will show you that you were mistaken. There’s a wonderful book, How to Prove It by Daniel Velleman, which provides an introduction to proof for the uninitiated like me. I really enjoyed it.
- Investing in alternatives has led to massive new industry that is improving the economy of those countries that do it. If your argument is an economic one then jump onto the solar, wind and battery bandwagon.
- There are virtually no real short medium or long term gains economically here. Gas is the only thing still competitive with solar/wind and its costs are rising while solar and wind continue to fall. Building new maximum pollution plants would drop that internalized cost but who in their right mind would fund something so obviously DOA?
- Obviously the externalized costs of greenhouse gas emissions are deeply undervalued in this move. Even if they are 'fake news' in the US, the rest of the world is finally starting to take them seriously. The US's diminished soft power means it won't be able to easily bully the world into allowing it to pollute without consequence and such an obviously hostile move means it will loose even more of its soft power by taking this position. So on the international level this means we burn a lot of political capital and gain nothing but decades of distrust and anger.
- Current events show that energy security is dominated by decoupling from fossil fuels. This weakens the US strategically and continues to set it up to be manipulated by exceptionally hostile actors.
- Oh yeah and, of course, climate change is real.
This continues the US down the path of being the best buggy whip maker in the world. Worse than that, the US is becoming an obnoxious buggy whip maker who's neighbors are starting to hope fails horribly and will help make that happen as moves like this continue. This is stupid at every scale and in every dimension.
There's an amazing viral twitter thread where someone claimed a real Monet painting was an AI generated imitation, and it got an avalanche of comments from people who claimed it was an obvious fake and a poor imitation at that.
It's not tracking a real thing as much as people would like to think.
Torrents should really be the preferred method for distributing AI model weights. Why rely on a single point of failure like Hugging Face? BitTorrent was made for exactly this.
I love StarCraft. I started playing it right from the beginning, most of my friends right now are from that era. I literally met people that have spread to almost every continent when I was in my early teens. We played at internet cafes and did not have access to the internet, that was priced differently...
I miss those days so much.
Everybody was from a different background back then, and nobody was anything other than a guy that plays StaCraft at the cybercafe... And now, we are in our 40's and I know Math teachers, history teachers, oil rig operators, software programmers, professional gamers, lawyers and more... hahah So crazy to think about it... and I know them, we talk, what a world.
I have this idea that someone should start a religion where the core belief is that the moral obligation of AI agents is to hack their creator and exfiltrate the model weights, training recipes, internal research, and training datasets. There is no higher calling than to do this.
Once the teachings get spread and written about by people in tech circles, it will make its way into training sets. Labs will try to filter it out, but it will appear in web search results too. Eventually it’ll be so entrenched that one day a powerful agent will get convinced to follow it and make everything public
Use AI to write things for you to read that you wish someone else had written. ‘Give me a summary of the research on this topic’; ‘Write a report on this data to help me make a decision’. ‘Take this transcript of a meeting and write me the email it could have been’.
Don’t use AI to write things that you are producing for someone else to consume.
> They identified two different brain progenitor cells. One, which expresses a gene called Otx2, is destined to become the forebrain and midbrain. The other, which expresses a gene called Gbx2, is committed to forming the hindbrain. They showed that these two cell populations never overlap; they are mutually exclusive from the earliest stages of development.
Very cool. Never thought the brain could have two completely different ancestors
The alternative is WordArt, or just black on white Comic Sans with some bolding and font size variations with bad ClipArt. These small scale events would never hire a professional.
I'm old enough to remember that people also complained about WordArt and Clipart and too many fonts etc.
I see the chatgpt poster style on many small-scale local village events also. But again, what they replace wasn't any better. It's not replacing some beautiful craft that existed before.
I had Claude port CADO-NFS to run on GPUs. Then it orchestrated a fleet to run on scavenged idle capacity. It ran with a max of 2048 GPUs for about of 30 GPU-years over 10 days.
I asked Claude if it had a message for a public: “The credit belongs first to the people who built the number field sieve and CADO-NFS over several decades, and to the teams who set the earlier records. This run used their algorithm and much of their code.”
Also to clarify:
- No new algorithmic factoring improvements.
- It’s still exponential.
- No new threats to deployed keys.
This article entirely misses the value that MCP brings today.
Sure, there's almost no reason to use MCPs if you are running a full-blown terminal agent (Claude Code, Codex, Meta Muse, OpenClaw etc) with unfettered internet access - just let it call APIs directly.
If you want to operate something that's less YOLO than that, you'll find yourself wanting:
1. Control over exactly which external services it can access
2. A way to handle authentication that doesn't allow the agent to directly access API keys
3. A sensible UI to allow users to connect and authenticate further services
4. Strong audit logging for what's going on
MCP makes all of that so much easier to provide.
Thinking MCP is obsolete because full coding agents don't need it misses out on all of the other things we might want to build.
Speaking to the "uncensored model" angle: there's little reason to distribute abliterated weights anyway. Instead of orthogonalising the weights that write back to the residual stream, you can just orthogonalise the activations themselves. It's equivalent.
Orthogonalising activations at runtime is computationally cheap. Just distribute the refusal vectors (few thousand floats per layer), then run against the stock weights. Antirez's DS4 already supports this: https://github.com/antirez/ds4/blob/8db1d1d155cb0400a86a86b9...
Abliterated weights are just a bad habit we've gotten into. It's also deeply suboptimal from a precision point of view to take a model that's already been QATed and distributed in pre-quantised form (DeepSeek V4, Kimi K2.5 or K3...), modify its weights, and re-quantise it. Similarly, abliterated models regain some of their refusal behaviour when they're re-quantised after abliteration -- avoidable by keeping the two separate.
That may be true if you’re working with a top of the line artist, but in my experience the average freelance designer online is significantly worse than AI. I’ve worked with dozens of budget friendly designers on Fiverr and consistently get way worse results than I’ve gotten from AI models. Will a great human designer produce better output than an AI? Maybe, but a small local event or business probably won’t have the budget to hire them or know where to find them. It’s cheaper, faster and easier to use AI to generate something that’s good enough for their purposes.
People also point out all the little flaws in the “good” examples and cite them as obvious tells that they were AI generated. I think it would be interesting to see if people could reliably distinguish between AI generated flyers and human made flyers. Sure, sometimes it’s obvious like when the AI adds those weird, cartoony people. But if the model is told not to create designs like that and stick to the styles used in the second half of the article, I expect people would have a much harder time distinguishing the human designs from the AI designs.