I would love for things like this to be accessible via hubs like open router or AWS bedrock. It's hard to justify adding new model vendors directly with all the heightened concerns about privacy and security, but if bold new capabilities are added to a centralized already-vendor like AWS, technical people can adopt them without going through a whole compliance/purchasing/vendor review process. And an extra middleman tax is well worth it when the cost savings of the model itself can be one-two orders of magnitude.
You can set privacy requirements and define an allow list. To me the main value prop is that I get one bill for all models and can quickly try new models without signing up anywhere or changing my code.
Oh! Also you can pass an array of models and if the first provider is down it automatically falls through to the next provider. More useful than it should be...
That's still ultimately privacy by contract (where you have to trust the inference providers to uphold their end of the deal), rather than privacy by design.
I think the trouble is that Typesafe APIs don't fit into the normal OpenAI-style API that every other regular LLM provider users. You're not just providing unstructured text and getting unstructured text back. It would take a different request and response format than every other model on Open Router. Though you could shoe-horn it in some way, it'd be hacky.
But agreed it'd be very useful to see it deployed on other hubs, and it seems worth it to provide the bespoke API format. Perhaps Typesafe's API will end up becoming the standard for a new type of structured model, the way OpenAI's API did.
It's really useful to have a chat app that is gatekept behind a 20k/year+ subscription. It's a particularly useful signal if someone is offering you a million+ dollar asset for sale.
I get that, was more curious if you knew how much that is used for signaling and the traffic there. I sort of expected these communities to exist in other sources (like regular SMS / Slack / private back channels etc).
I'm going to copy-paste one of the comments from ryukoposting higher up in the thread because I think it's relevant here:
> 2) Bloomberg's chat is important because, as a hedge fund or investment bank, the chat is how you buy and sell bonds. You agree to the trade in the chat, then tell the back office folks to execute the trade. Direct quote: "I'd wager 90% of the ~400 trillion in annual bond trading value happens over Bloomberg DM"
You may discuss trades via SMS or Slack; maybe you're important enough that you have a broker, and you send them your idea. But the broker at the investment bank is probably handling those trades via Blomberg terminal. That's the network effect: all the bankers are in Bloomberg chat, and it is therefore the easiest place to find who owns an asset and negotiate a trade.
Xi Jinping rose to power on a message of anti-corruption, and part of the reason he remains in power on an indefinite term is by presenting himself as the "only trusted" person to maintain anti-corruption amongst all the factions.
While I'm sure he doesn't catch all corruption and the CCP overall has selective enforcement, the reason they do have measures like this is in large part because of Xi Jinping's specific reputation and positioning.
Or in authoritarian regimes it often means "stole from the party". As long as you only steal from the poor and give the proper bribes up the chain you are in no danger in most corrupt societies. Except possibly in the rare occasion where your corruption causes a disaster that embarrasses the people above you in the hierarchy.
Far too often when you see stories about how someone was persecuted for corruption it boils down to "he stole from rich people".
This is true in any power structure ever. Kings, mafia, pick-your-dictatorship, many democracies. Hurt or steal from the poor, not other rich or elite, and always make sure to kick some up to the big guys.
All presidents rise to power on a message of anti-corruption, they all clean house when they start in office but usually it falls off until the next president takes over. The problem Xi has is that now that he is now president for life, the house isn't getting cleaned in the usual way every 10 years; he has to do a corruption purge every so often or things will get grimey.
I'm just talking about China, we were only talking about China right? We don't really have a lot of data points to go by since there have been only three presidents so far with supreme power (after it was combined with General Secretary and military chair head, before that president was more of a ceremonial role).
This looks like a good approach, though I would expect this to be a native macOs feature within 12 months -- this seems totally like it fits into their product roadmap.
I looked for this in the past. This is the main reason we bothered with mailchimp/hubspot -- simply the ability for nontechnical marketing people to put together nice emails, and the trust that we won't need an engineer to troubleshoot email formatting on their behalf. I remember trying some OSS tools at the time (8 years ago?) and there were some templates we used but then when we wanted to modify them, the broken-ness of email html/css standards made it really hard to test.
I know the standards and practice around this are a moving target, though, so I hope you can find a model to sustain and expand this, without charging for delivery/contact list numbers like MailChimp or other incumbents.
Thanks — this is exactly the audience I built it for.
The cross-client rendering thing is why Templatical outputs MJML instead of raw HTML. MJML was built to abstract all the
table-based, Outlook-2007-quirks, Gmail-strips-style nonsense — you write semantic blocks, MJML compiles to table HTML
that works across every major email client. So when your marketing person moves a block or changes a button color, it
doesn't silently break in Outlook two weeks later.
On sustainability — same concern. Even while I was building it, multiple times I caught myself asking "is this even worth theeffort? Maybe not with all the functionalities I've built, but someone could vibe-code a lightweight version of it in a day." But at the same time, I see and personally used SaaS products with the same or fewer features selling for $2,500/mo, which seems ridiculous.
I'm currently working on a subscription-based Cloud version, but only for things that actually need an infrastructure and backend: AI chat/rewrite, image-to-template conversion, MCP integration, hosted media gallery, saved modules, commenting, real-time collab, email testing, version history, etc. Sending stays your own provider — no per-contact, per-email, or per-delivery charges.
Even when using mjml you must do render tests of your templates using litmus or a similar service. Mjml is WAY better than hand rolling templates but there are foot guns to be found and the docs won’t mention most of them https://github.com/mjmlio/mjml/issues/2927#issuecomment-2539....
Totally agree on edge case quirks of MJML - or more like weird behaviors in Outlook. With so many email services and clients around, nothing can provide 100% bulletproof solution but I think MJML is as close as it gets to works-for-majority. I‘m sure MJML maintainers also aim for this.
Also, yes! Microsoft should kill Outlook with fire! :)
Is it really a mistake? OpenAI's own agent SDK also has a Claude.md file. That's not an indication that OpenAI internally use Claude, rather, it's there because the SDK has multi-model support.
I used to help nonprofits and small businesses build websites. Process always went like 1. buy domain, 2. buy a shared hosting provider that one-click-installs Wordpress, 3. use a theme to begin editing the website. Often, I would also use the email included with that hosting provider for the firm.
ALL of that goes through cpanel, for every shared hosting provider I can ever remember using. Even if the stuff happening on those servers didn't use perl, cpanel itself -- the admin of everything provided for that domain by the hosting provider -- it's a huge surface area.
reply