matznerd

If you are a company and can't figure out how to improve your bottom line through increased per person productvity, sales, product development, even just responses to customers etc, then that seems kind of like a "you" problem and not problem of the AI models...

I think you can make it work without leaving Gists by using a Cloudflare Worker as a workaround. The Worker sits in front of the renderer page and adds the og tags to the HTML before it's sent out. You'd also need to turn the SVGs in the Gist into a PNG for the og:image, and decide if you want a grid or just one image, any text formatting, and how long to cache...

I got it working in a quick local test (grid of all the reasoning efforts, cached per Gist, loads from the raw Gist URL so it doesn't hit the GitHub API rate limit).

Code + prompt + notes here: https://gist.github.com/matznerd/ece297107bd99ac028c7962c217...

Basic concept is to:

1. Put a Worker on the /markdown-svg-renderer route. Normal visitors get your page exactly as it is now.

2. When a link has ?url=<gist>, the Worker reads the Gist and adds og:title, og:description and og:image to the page's HTML. Link previewers like Slack and iMessage don't run JS, so this is the only way they see them.

3. og:image points to a second Worker URL (og.png?url=<gist>). It takes the SVGs from the Gist, puts them in a grid, and converts it to a PNG, since previewers won't show SVGs.

4. Both results get cached per Gist, so each Gist is only fetched and rendered once, even with a lot of traffic.

Things to customize:

- Title and description (mine: "gpt-6-luna SVG of a pelican riding a bicycle" / "6 runs, reasoning effort none to max")

- Grid of all runs vs just one image, plus layout, labels and font

- How long to cache (I used a day, but edited Gists keep the old preview until it expires)

Simon, love your work, one piece of minor feedback for the individual model pages is to make the font of the model name potentially bigger than (and above) the conversation id (which means nothing to the audience) "2026-09-22T18:28:00 conversation: 01m355zvyw8946qyraa8zpz6h9 id: 01m355zvyx47zxx5c6q6b3fg0m#".

I had all the tabs open individually and harder to scan which model is which... otherwise keep up the great work! I like the grid view a lot. (Also the pages have no OG images set, which impacts what the link looks like shared)...

Not OP, but you can't set a per api key limit in OpenAI Platform, you also can't really easily see spending going on per key, and its also in a far off page separate from the keys, really not easy to see what is going on with spend in real time.

Is there any way to make it so that when you plug new things in, the power doesn't renegotiate causing other things plugged in to reconnect/flicker/interrupt.

Looking for high power USB C hubs that don't do that, but seems most do as part of PD and need isolated slots that don't draw from shared potential... programmatically locking specific power to laptop hub, while others negotiate would be ideal...

Great that it was available, but compared to Peter's version this one is inferior. It didn't have draft email as a default and asking it to write a draft would just send the email, oops lol. And doesn't have mutli-account support or a number of other features. I think one thing it could do better was inline commenting (maybe), but neither CLI can initiate their own comments...

Gog cli - https://github.com/openclaw/gogcli

have enough sessions going that you just cycled through them and by the time you get back that chat is ready to go again. Easier sometimes staying in same project, sometimes easier if you don't use worktrees to work on different projects.

I thought this post was going to be a metaphor about how most people can barely handle 1 project, while some people need to multiple projects for it to feel natural...

I do writing with RAG and it can be implemented to suprisingly good if you already have your own writing that the text is being generated from. FAQs etc can be pretty easy when your content is context for the AI.

After a few rounds of AI generating AI content from AI content, I'm sure it could eventually become slop...like the model collapse lol idk.

"AI models collapse when trained on recursively generated data" - https://www.nature.com/articles/s41586-024-07566-y

Dear Notion employees, please, if you're advertise mail and calendar as features, add them into your app and do not make them open in new tabs. I want an all in one thing, why is that so hard, you already have tabs built in? Thanks!

okay but what about "c" being nearly the same as "z", neither of which look like the character and are nearly(?) identical. Is our brain supposed to just be able to figure it out?

"A large new study published in the International Journal of Hygiene and Environmental Health provides evidence that exposure to certain workplace chemicals among parents may influence the severity of autism spectrum disorder (ASD) symptoms and contribute to behavioral, cognitive, and adaptive challenges in their children. The findings suggest that occupational exposures—especially to plastics, ethylene oxide, phenols, and pharmaceutical agents—may have broader developmental effects beyond autism diagnosis alone."

"The effects of parental occupational exposures on autism spectrum disorder severity and skills in cognitive and adaptive domains in children with autism spectrum disorder" https://www.sciencedirect.com/science/article/pii/S143846392...

The person leading this study, Erin C. McCanlies, was forced out of the CDC, her division eliminated and she went into early retirement from the CDC. https://www.psypost.org/scientist-who-linked-autism-to-chemi...

---

"The findings suggest that workplace exposures to several specific chemical classes were associated with worse outcomes in children with ASD. One of the strongest and most consistent patterns involved plastics and polymer chemicals. Fathers’ exposure to plastics was associated with lower scores across all cognitive and adaptive skill domains, including language, motor coordination, daily living skills, and overall functioning. When both parents were exposed, the deficits appeared to compound.

“I was surprised how strongly and consistently plastics and polymers stood out as being linked with multiple developmental and behavioral outcomes including irritability, hyperactivity, and daily living,” McCanlies told PsyPost.

Exposure to ethylene oxide—commonly used in hospital sterilization—was also linked to more severe autism symptoms, lower expressive language abilities, and poorer adaptive functioning. Similarly, parental exposure to phenol (used in construction, automotive, and some consumer products) and pharmaceuticals was associated with increased ASD severity and more pronounced behavioral challenges, especially hyperactivity and stereotyped behavior.

While the results do not imply that all children exposed to these chemicals will develop more severe symptoms, the patterns suggest that early life exposure to workplace toxicants may amplify certain developmental difficulties in children who already meet criteria for ASD. The study provides one of the most detailed looks to date at how parental occupation may relate not just to diagnosis, but to variation in how autism is expressed.

“Our findings suggest that certain parental workplace exposures may be related not just to autism, but to worse symptoms and autism behaviors,” McCanlies explained."

hard disagree

These are 8bit versions On mac, use LM Studio to download it, just search "oss mlx", and note there is an mlx toggle box on search.

Link for the 120B version: https://huggingface.co/lmstudio-community/gpt-oss-120b-MLX-8...

Its taking 21 gb of memory on my 64 gb mbp, still tuning it and settling on context size, temp, and other settings.

My comment from yesterday:

"thanks openai for being open ;) Surprised there are no official MLX versions and only one mention of MLX in this thread. MLX basically converst the models to take advntage of mac unified memory for 2-5x increase in power, enabling macs to run what would otherwise take expensive gpus (within limits). So FYI to any one on mac, the easiest way to run these models right now is using LM Studio (https://lmstudio.ai/), its free. You just search for the model, usually 3rd party groups mlx-community or lmstudio-community have mlx versions within a day or 2 of releases. I go for the 8-bit quantizations (4-bit faster, but quality drops). You can also convert to mlx yourself...

Once you have it running on LM studio, you can chat there in their chat interface, or you can run it through api that defaults to http://127.0.0.1:1234

You can run multiple models that hot swap and load instantly and switch between them etc.

Its surpassingly easy, and fun.There are actually a lot of cool niche models comings out, like this tiny high-quality search model released today as well (and who released official mlx version) https://huggingface.co/Intelligent-Internet/II-Search-4B

Other fun ones are gemma 3n which is model multi-modal, larger one that is actually solid model but takes more memory is the new Qwen3 30b A3B (coder and instruct), Pixtral (mixtral vision with full resolution images), etc. Look forward to playing with this model and see how it compares."

Also, these small models are meant to be run local so not going to appear on openrouter...

Here are the LM Studio MLX models:

LM Studio community: 20b: bhttps://huggingface.co/lmstudio-community/gpt-oss-20b-MLX-8b... 120b: https://huggingface.co/lmstudio-community/gpt-oss-120b-MLX-8...

thanks openai for being open ;) Surprised there are no official MLX versions and only one mention of MLX in this thread. MLX basically converst the models to take advntage of mac unified memory for 2-5x increase in power, enabling macs to run what would otherwise take expensive gpus (within limits).

So FYI to any one on mac, the easiest way to run these models right now is using LM Studio (https://lmstudio.ai/), its free. You just search for the model, usually 3rd party groups mlx-community or lmstudio-community have mlx versions within a day or 2 of releases. I go for the 8-bit quantizations (4-bit faster, but quality drops). You can also convert to mlx yourself...

Once you have it running on LM studio, you can chat there in their chat interface, or you can run it through api that defaults to http://127.0.0.1:1234

You can run multiple models that hot swap and load instantly and switch between them etc.

Its surpassingly easy, and fun.There are actually a lot of cool niche models comings out, like this tiny high-quality search model released today as well (and who released official mlx version) https://huggingface.co/Intelligent-Internet/II-Search-4B

Other fun ones are gemma 3n which is model multi-modal, larger one that is actually solid model but takes more memory is the new Qwen3 30b A3B (coder and instruct), Pixtral (mixtral vision with full resolution images), etc. Look forward to playing with this model and see how it compares.

Love the work you're doing at Unsloth!

If you're on Mac you can download LM Studio and get the MLX version (edit - links below). I am running it on 64 gb M1 and it takes about ~30 gb ram. I've been on the hunt for a local orchestrator model that interprets input with speech to text (STT) from WhisperX, then can decide what to do. I have only been running it for a day, but it may be overkill for my setup.

For simple tasks it can quickly respond and then understand to use MCP servers for tasks and other things, but offloading all the heavy lifting to claude code via sdk and cli, then bringing the results back in a summary or with clarifying questions as text to speech (TTS). I'm playing with Kyutai TTs b/c have great models that sound real and can do conversational streaming with VAD (though my mbp is too slow with it for now but see https://unmute.sh/ for demo).

I am looking for an orchestrator model that runs on 10-15 gb of ram and can do really good tool calling and model routing. I'm will likely move to something even smaller designed specifically for this, like Jan Nano and then spin up an intermediate model like Qwen if needed, or try a smaller Qwen. https://github.com/menloresearch/jan?tab=readme-ov-file

Ultimately, I want something that can see my screen and know what is going on and have full context and be live, so I was excited about Gemma 3N multi-modal, but its not really available yet fully with vision at least for MLX. https://deepmind.google/models/gemma/gemma-3n/

Next 6 months in this area is going to be pretty wild though.

edit: Fixed links below, thanks

https://huggingface.co/lmstudio-community/Qwen3-Coder-30B-A3... https://huggingface.co/mlx-community/Qwen3-Coder-30B-A3B-Ins...

go to chrome://flags ,

follow these instructions https://youtu.be/q7dnkGdndNo?t=220

then load extension in developer mode

I think what you're missing is that this is likely in preparation for some sort of AR glasses and something smaller form factor than the Vision Pro that will require transparency so you can see through to the real world etc...

There is a reason that longevists, longevity researchers, and biohackers focus on mitochondrial enhancement through various means, supplementing NAD/NR, Ubiquinol, PQQ, ALCAR, photobiomodulation, etc

"Your eyes transform light into electrical impulses that coalesce into an image in your visual field, and your ears transform air-pressure waves into electrical pulses that you eventually perceive as sounds. Likewise, mitochondria transform dozens of hormonal, metabolic, chemical, and other information streams into their electrical membrane potential. This “bioenergetic” state then leads to the production of secondary messenger molecules that are intelligible to the nucleus. So in the same way you read messages on your phone, which receives signals, transforms them and projects decipherable information onto its screen, the nucleus of your cells can “read” the environment through the MIPS that surrounds it. Rather than having supplementary roles like those of battery chargers, mitochondria are more like the motherboard of the cell. Genes sit inert in the nucleus until energy and the right message come along to turn some of them on and some others off. Mitochondria provide these messages, speaking the language of the epigenome—the malleable layer of regulation that sits on top of the genome to regulate its expression.

My colleague Timothy Shutt of the University of Calgary likes to call mitochondria the “CEO of the cell”: the chief executive organelle. This metaphor captures how mitochondria not only are involved in integrating information but also give orders. They dictate whether the cell divides, differentiates or dies. Indeed, mitochondria have a veto on cell life or death. If the MIPS deems it necessary, it triggers programmed cell death, or apoptosis—a form of self-sacrifice for the greater good of the organism.

So vital are mitochondria that in difficult times cells may donate entire mitochondria to other cells. “In cellular emergencies, newly arrived mitochondria might kick-start tissue repair, fire up the immune system or rescue distressed cells from death,” journalist Gemma Conroy noted in a Nature news story last April. Inside tumors, cancer cells and immune cells appear to compete for mitochondria, using them as a kind of bioweapon. An international effort I participated in, led by Jonathan R. Brestoff of the Washington University School of Medicine in St. Louis, recently created an entirely new lexicon to guide the emerging field of mitochondria transfer and transplantation. All well and good, you may think. What does all this mean for my health or how long I’m going to live?

The short answer is that it may have everything to do with human health. Diabetes, neurodegenerative conditions, cancer and even mental health illnesses are all emerging as metabolic disorders involving malfunctioning mitochondria. And these findings are indicating new routes for intervention.

Mitochondria drive health—or disease—in several ways. One route derives from their role as energy processors. In an electrical circuit, if we crank up the input voltage too much, we can blow it out. Similarly, if our cells are exposed to too much glucose or fat—or, worse, both together, causing what doctors refer to as glucolipotoxicity—the mitochondria undergo fission and fragment into little bits, accumulate mtDNA defects, and produce signals that end up prematurely aging or killing the cell. Experiments in cells and in mice have shown that pharmacologically or genetically preventing mitochondrial fission induced by excessive glucose and fats may protect against insulin resistance. Cancer, too, may be a disorder of cellular metabolism. Cancer cells can burn glucose without oxygen, which suggests either that something is wrong with their mitochondria or that they prefer to reserve mitochondria for use in cell division—and proliferation.

A second pathway is through mitochondria’s influence on gene expression. Mitochondrial signals alter the expression of more than 66 percent of genes in the nuclear chromosomes. By changing which genes are expressed and to what extent, mutations in mtDNA may completely alter the nature, behavior and stress resilience of cells and ultimately of the whole organism."

What about "fast apps" as in apps you build with AI to quickly fill a niche knowing it won't be a long term viable business, but build to just for that moment?

You're missing that Claude desktop has MCP servers, which can extend it to do a lot more, including much better real life "out of the box" uses. You can do things like use Obsidian as a filesystem or connect to local databases to really extend the abilities. You can also read and write to github directly and bring in all sorts of other tools.

This is one of the ways we get all of the secondary metabolites from plants, most of the compounds in supplements and drugs that are active in humans. They come from mutations of the existing DNA, then are able to change from there without impacting the plant's core functions due to being duplicates:

"Gene duplications drive the recruitment of genes for secondary metabolism. Gene copies are gradually modified to create genes with specificities and expression patterns adapted to the needs of the new pathway in which they are involved. Duplicated genes are often in tandem repeats, forming clusters within the plant genome. However, in some cases, clusters of nonhomologous genes have also been identified as forming a functional unit. The selective forces that have caused the establishment of new pathways are far from understood and might have changed repeatedly during evolution owing to the continuously changing environment. Recent data show that the way several classes of secondary compounds are scattered among species is attributable to independent recruitment and the inactivation of biosynthetic enzymes."

https://www.cell.com/trends/plant-science/abstract/S1360-138...

If you want to understand secondary metabolites, which is not the "how" do they work part of plant-derived molecules, but the "why" do they work?

You can follow this line of research:

"Over recent years, the consensus as to the mechanisms responsible for these effects in humans has shifted away from polyphenols having direct antioxidant effects and toward their modulation of cellular signal transduction pathways. To date, little consideration has been given to the question of why, rather than how, these plant-derived chemicals might exert these effects. Therefore, this review summarizes the evidence suggesting that polyphenols beneficially affect human brain function and describes the current mechanistic hypotheses explaining these effects. It then goes on to describe the ecologic roles and potential endogenous signaling functions that these ubiquitous phytochemicals play within their home plant and discusses whether these functions drive their beneficial effects in humans via a process of “cross-kingdom” signaling predicated on the many conserved similarities in plant, microbial, and human cellular signal transduction pathways."

https://www.sciencedirect.com/science/article/pii/S216183132...

The functional aspect of the duplicated DNA goes back to the point that humans, insects, and plants are all eukaryotes...

How can we trust a human to run the study, isn't there a bias? Needs an AI prompted to be a research scientist as a co-author for it to be balanced.

A better question is how do we not have a dark mode/system settings for HN?

Hi Boris et al, can you comment on increased conversation lengths or limits through the UI? I didn't see that mentioned in the blog post, but it is a continued major concern of $20/month Claude.ai users. Is this an issue that should be fixed now or still waiting on a larger deployment via Amazon or something? If not now, when can users expect the conversation length limitations will be increased?

Lol I don't think there are that many, which area? But gas stations can be like normal businesses there and often have restaurants and sometimes even hotels. I accidentally stayed at a gas station hotel once by the airport, when my go to hotel was full and I didn't know and booked online, and only when I arrived did I realize it was a gas station.