sim04ful

Sol and 5.5 pro are in parity at $5 input / $30 output. What I'm inferring from this is that: - model weight size didn't change, and this is mostly a result of better model architecture and scaled up RL - better hardware utilization and and they're making better margins OR - worse hardware utilization and they're okay with digging into their margins.

"We're also launching GPT‑5.6 Sol on Cerebras at up to 750 tokens per second in July, bringing frontier intelligence to customers at unprecedented speed. Access will initially be limited to select customers as we expand capacity."

This seems like it would be the largest and first closed-source model Cerebras has offered till date

I simply cannot see how crypto can go mainstream without first solving the on-ramp experience.

I certainly hope so, it's cheaper than ever to write native code, but if sub-trillion dollar valuated companies like OpenAI have done the math and still use electron for the codex app, one has to ponder what was considered.

I notice that the concept related to the right answer sometimes has an opposite counterpart.

Awesome project, love how straight to the point your landing page is. I was building something similar to linkedrecords but on the internet computer (https://internetcomputer.org), i went about things naively compared to this, and decided working on the problem wasn't worth as much effort as completely reformulating it (different topic)

Yeah i feel separating compute from data is one of the best directions the industry can move towards, it means less lock-in, more flexibility and user agency as well as more personal compute utilisation (increasingly important with the cloud compute pricing-in the RAM supply crisis).

Could i ask, what made you decide to work on this ? personal and/or enterprise ?

I’m building https://design.withfudge.com, a Prolog-backed design search engine that lets designers/agents query structured design knowledge from real websites. It uses data from my other startup, https://fontofweb.com, to help designers find concrete inspiration e.g fonts, colors, layouts, screenshots, and patterns, so they can make better design decisions.

I'm rather surprised it took Meta this long in the wake of X introducing subscription plans quite long ago now.

"The api pricing for mimo-v2-pro and mimo-v2-omni remain unchanged" could we presume this means the discount isn't from hardware improvement or availability ?

There's also https://internetcomputer.org/

My opinion is that we're using the wrong paradigms for LLMs. We should be leaning more on declaratively specifying behaviour.

If there's any hope for reliability, auditability, predictability to be had it lies in contraining and LlMs grammar whilst delegating freeform behavior to a more passive substrate.

On a related note they opensourced an email client: https://github.com/cloudflare/agentic-inbox

For advice let me use my current product. https://fontofweb.com is essentially Pinterest for web design. It does semantic search against a database of UI screenshots and recordings.

The problem with products like this is that for search to be useful, you need, well... data. And when your data is crowdsourced, you need, well... a crowd. That was the chicken-and-egg situation I found myself in.

So the way I solved it was by painstakingly clipping and recording ~8500 websites myself.

But the great thing about that arduous process is that it forces you to put yourself in the shoes of your typical user. And really, the only way to make good products is to have a deep level of empathy for that user. In my case, that meant streamlining the data creation process itself, the act of pinning a screenshot or recording, because I had to do it over and over again. Making it easier for myself ended up making it better for everyone else too.

Of course, all that data-seeding effort goes to waste if search engines and potential users cannot discover it. So you have to go all in on SEO, especially sitemaps, and on programmatic socials where possible. This is especially true for a product like mine.

A lot of these marketplaces also have relatively homogenous data, which is actually useful, because it means you can present it through templates and publish pages at scale automatically.

Graphql introspection queries would be a really neat application for LLM calls

MCP server to give AI agents design taste - https://fontofweb.com/mcp

Agents can search for design inspiration from production websites using semantic search. Since this inspiration comes from live websites, their design tokens; colors, typography usage, layout data are also available.

This seems like it would have adversarial consequences. Wouldn't these list of tropes get longer over the years.

I've noticed a key quality signal with LLM coding is an LOC growth rate that tapers off or even turns negative.

Nothing on symbolic reasoning ?

Looks pretty interesting. How could i use this on other MCP clients e.g OpenCode ?

I daresay we're going to see a burgeoning situations where the software (code) is open-sourced under a permissive or copyleft license, while the associated data, content, or assets (e.g., datasets, models, or databases) are handled under separate, often more restrictive licenses.

I'm also working on a Chinese learning app (heyzima.com) and my "solution" to this was to use the TTS token/word log probabilities.

The prevalence of this "personal vibecoded app" spirit makes me start to wonder if an "App" is the right level of abstraction for packaging capabilities. Perhaps we need something more "granular".

I'm very green to this so forgive if this question sounds silly:

Would instead of the RL step a constrained decoding say via something like xgrammar fix syntax generation issue ?

This is a godsend for me, thank you.

Not a stranger but strangers I was returning home from an event early evening. Being absorbed in my thoughts. I got both my front tires free spinning without traction in a ditch.

Although this was in Nigeria, we have this certain camaraderie through hardship, it was still extremely surprising seeing a group of 6 men come out of nowhere, having nothing to do with each other aside being passerbys join hands, exerting sweaty effort to get my car out a ditch by 8pm.

Left me quite an impression

The main issue I'm facing with realtime responses (speech output) is how to separate non-diegetic outputs (e.g thinking, structured outputs) from outputs meant to be heard by the end user.

I'm curious how anyone has solved this

What's 09h/16h ?

LLM spam ? on my HN ?

It's actually funny, the lack of styling actually makes consuming the info alot easier.

I created a free collection of 4,300+ real website designs (screenshots, fonts, colors, live links)

https://fontofweb.com/u/fontofweb

I was tired of inspiration sites like Dribbble full of polished mockups that aren't practical. Or awwward like sites that don't represent the mundanity of most websites.

So, I spent a while building a tool that captures website design snippets. It's now a collection of 4,363 designs from 544 different domains.

For every design, it extracts:

The exact fonts used on the page (so far 561 unique font families I've found)

The precise color palette

A direct link to the live site

You can check out the full free collection here: https://fontofweb.com/u/fontofweb