dzink

The content within the models might be the play. If inserting the right content for the rest of the world to consume from the models is important to them, they will give away all the content they want the world to have.

And yet the Mata price jumped yesterday. For some investors this is apparently a feature.

Just read Careless People. It’s good and this triggers the Streisand effect.

When everyone uses AI to study the same indicators and figures out how the prices move with those indicators they all start investing at the same time and the prices move together. AI silently gets everyone on the same page.

Without a large funded team the founder who knows what they are building will not build an expansion line into other areas because they can’t support the growth solo. Somebody once said funding is kerosene. You want to pour it into working fire pits, not sparklers hoping one of them will be a dynamite stick.

Here is another story: A baker who bakes for her kids every day makes an oven. She spends years perfecting it by herself with details only someone who uses this product would notice. The nuance of gold baked details in just the right places on the bread, the infusion of essences for the cakes. The precise charring on the pizzas. She goes to young founder events to meet likeminded makers and they talk about space ovens and OvenCrunch incubators fund them just on school name and ideas. But the oven maker with the kids doesn’t even get an interview. She applies with a working product and increasing sales year over year for 10 years and no interview. Her over becomes an organically growing best seller and she doesn’t need the seed money anymore. Incubator founders have spend their seed funding on fancy trips and conferences and flying over the Egyptian pyramids on instagram. The Incubator partners say they don’t fund oven makers anymore because the business is too slow to grow and consumer stuff is a tarpit.

We are 81 years away from the end of World War II. The baby boomers born after are the ripest target of financial sharks aiming to get a chunk of their retirements. People will be liquidating to settle estates, to pay inheritance taxes in large numbers. The AI boom feels like a stealthy rug-pull from other assets that are likely to tank from retirement withdrawals and into something that may last a little past the boomer assets wave.

Look up online wishlist. An online tool that people will not find, because the results are articles instead of products in that category. Look up dreamlist. You won’t find it either, even though it is another private tool that lets families and gift drives create collaborative lists for disaster recovery and the holidays. Small search engines need to realize Google has an advantage with chrome because it counts duration of stay on sites, instead of just content and can tell if something is a tool (people spend more time on those).

Or you can do a much cheaper tiny esp32 typewriter with a basic display and a usb for a keyboard. But who wants cheap for education.

If I build you a custom device that receives questions from a central computer and lets the user plug in whatever keyboard they want via bluetooth or usb to only type and answer the questions (ability to edit the text and submit when ready). The central computer can receive all the questions submitted to any of these devices connected to it via wifi or cloud. How much would pay per device? Would you pay for a subscription?

Fill the room with typewriters.

The goal should be to not repeat Star Wars - new technology that gives power and advantages is always adopted by companies first, then governments, and then it’s used to subjugate people. In Star Wars you have the trade federation abusing little planets. Little planet leaders go to government for help and vote a more “strong” approach to government that will “fight for the little guy”. An authoritarian leader steps in and uses the new technology to subjugate galaxies. In Foundation that leader can also clone himself, leading to 1000 years of fascism. Today prize camels and horses are being cloned in the middle east. Fiction is a guidebook sometimes.

Have you tried llama.cpp with unsloth and models suited to it? GLM flash? It seemed to allow more models to be tried soon after they are released. Haven’t tried for long term deployment though, that’s the next step.

Keep the Social, ditch the media.

Here is where we’ve found those rooms: 1. In scouts gatherings at churches or campgrounds (zero religious connotation - just use of unused space during the week). 2. In play dates with other families. 3. In sports clubs that have regular practice. 4. In local libraries which are a fantastic resource, especially for caregivers of young kids. 5 In local elementary schools where playgrounds are open and older kids can ride bikes in the yard. 6. In local parks. The market tries to capture the surplus capacity of people with no other obligations or kids. The hard part is finding the trustworthy people who will be a part of your community and the fellow families who want the same community for their children. Trust-based communities will carpool, and take turns for hosting play dates, and ignore messes to enable social interaction, and keep screen time off to ensure social interaction instead of zombie mode. There is market incentive to it and it is rare and it is hidden because it is rare and fodder for abuse. Trust networks exist in parallel to the market but they take offering value to receive value and not in a monetary way and they are also very unevenly distributed. The market profiting from gambling, and addiction, and alcohol means there are fewer safe places for this kind of network to build. But it can be started with as little as 2-3 families banding together.

Looks like a PR stunt sales pitch for their product.

I keep thinking about what humans will need in an age when AI keeps getting better. https://www.makeaplace.com (new product launch is imminent).

You’ve described a corporation. Think of YC companies - equity owned by early employees. Controlled by “someone senior”. And if things are going good more friends are invited to join in, multiple companies start owning equity within each other and you back new incomers by directing demand towards their products. The difference between that and a coop is that in a corporation there is some legal standing to enforce the rules and control abuse. In eastern europe where coops had more of a presence the people in power would just abuse the coops - theft, misdirection of resources towards “friends”, pyramid schemes benefiting “older members”… but you couldn’t go and sue as a shareholder because via political connections the entrenched leader and his buddies could use their power to pay off someone and hurt you. The real prof in concepts like this are enforceable contracts.

Visited a recycling facility recently. It’s a private company that covers an entire county in California. They filter the garbage wit people and a big machine and seem to get paid by companies abroad to ship them all recyclable materials - plastics, cardboard, metals, glass. That pays enough to keep them in business for decades. Someone really needs to look at where the materials from our garbage go.

If apple provides native support with enough bandwidth to run an external NVIDIA GPU for Inference and training, I will upgrade to the latest MBP instantly. Raise your hand if you would too.

The economist in me immediately asks: Where is the financial incentive to do this? Just the same way the programmer would ask what the stack is. Some possibilities:

1) Money laundering - large content farm someone can argue makes xyz in revenue to hide an alternate source of revenue.

2) Ad fraud - leading up podcast charts or SEO results to attract clicks to sell ads. Bot farms could also be making clicks to pretend sell ads as well.

3) Attempt to dominate the niche for sale of knitting products. Or to pretend to dominate it so they can sell their the business later at a larger multiple.

4) Test the waters of a much bigger engine for doing 1-3 above in an innocuous hidden subject, before they do it with elections or some other more profitable field. Regulatory waters as well - seeing what they can get away with.

Feel free to brainstorm more incentives for making something like this.

Nevada makes it much harder to sue corporate officers when they do malfeasance. Wyoming has tons of privacy perks for the officers (similar to cayman island accounts). “Perks” though also convert into signaling for the intent of the founders.

There is no insinuation. I am looking for both private and quality inference. I am not sure why Tinfoil had the characters. I will keep trying them, but it’s an issue if not resolved.

Tried DeepSeek V4 Pro and Flash on Open Router and they worked fine - flash might have actually produced a better result, but also the same prompt across different inference providers produced the same result. Then tried DS4 Pro again via tinfoil.sh and got the same design but littered with random Chinese characters in the code. Tinfoil pegs prompt data as private / not trained on. Do know know DS4 providers that are verifiably private and not training on your prompts and outputs?

My push for trying local was the wildly unpredictable but systematic performance of large models like Opus and ChatGPT. It feels like at different times of day or week they are getting degraded beyond belief. I don’t know if it is deliberate, a function of demand, or a function of the models themselves. We are all learning the shape of this space by trying. I need to be able to rely on consistent performance - and maybe that means putting some harness of benchmarks between models and maybe it means between different inference providers, and maybe local.

Children learn by playing because not much is expected of the outcome in play. Improvement happens when you can play. When AI has a play environment to learn with reinforcement. When entrepreneurs are allowed to try and fail and do better. Doctors learn by practicing under supervision, or on corpses, until they can do it for real. No straight line goes up without a jiggle in the beginning.

Go to Open Router, ask your own in investigative prompt that meets your needs to all the top open models. See how they do. Then notice if you can run any of those locally. Repeat at least once a month.

He wrote for aesthetics and he wrote for politics. In the end, he saw the aesthetic writing as meaningless.

Tiktok was launched by ByteDance in 2018. Reels was unleashed 2020 and YouTube shorts in 2021.

Claude has become practically unusable for Pro users in the past few days. The Opus 4.7 blew through an entire 5 hour limit in one question and didn’t even finish answering it. Zero value delivered.

Opus 4.6 is giving 2, maybe 3 questions before blowing through the Pro 5 hour limit as well. We are forced to use Sonnet which makes the same mistakes over and over and then to start trying with other companies. To make matters worse, it reuses old code as we try to survive between credit expiry so it re-introduced issues into the code with the limited credits, that we had already fixed on our own and with other models.

Anthropic in just a few days has gotten me to try GLM 5.1, the new Kimi, and back to OpenAI. OpenAI also seems to introduce new bugs without being carefully micromanaged. The advantage Claude has is that the models are more careful and can refactor code instead of leading to bloat as they go. But the throttling happening now is breaking things and making the entire subscription unusable. I really hope they fix it soon.