KC

Codex mode UI now pretty much exactly the same as Claude. Two modes: Home/Work and Coding. Place your bet on when two goes to one and the first to get there? UX (inc UI and other modalities e.g. voice) should wrap itself around the user's intent. You only need a single starting point, a place to begin that adaptation.

@LilBoyWander @openshipio Never tried self hosting, chiefly because this is exactly the kind of warning I have read before r.e. email.

I’ve been down this road, many of us have. Email is tough to have self hosted. New IPs have zero reputation and often land in spam permanently with Gmail, Outlook, etc. You need perfect rDNS, SPF, DKIM, DMARC, slow warming, and constant monitoring…that’s the hidden cost of something like this. One slip and everything dies in spam. I do not suggest doing this unless you have the time and patience to make it work. @resend is my go-to, along with Coolify.

Introducing OpenShip. An open-source application platform for building, deploying, operating, and scaling applications on infrastructure you own. Replace deployment tools, managed services, and infrastructure workflows with one open-source platform. Available today: - Mail Server (Built in. One click). • Runs on your own VPS • Unlimited domains • Unlimited mailboxes • Modern webmail included • Connect with Gmail, Outlook, Apple Mail, Thunderbird, or any IMAP/SMTP client • Send email directly from your applications using SMTP • No mailbox subscriptions or API fees • High email deliverability - Deployment: • Deploy any stack • Git-based deployments • Zero-downtime deployments • One-click rollbacks • Development, Staging, and Production environments • Multi-branch deployments with isolated environments • Deploy to VPSs, dedicated servers, cloud VMs, or your homelab - Services: Provision the services your applications need in one click. Replace multiple managed providers with services running on your own infrastructure. • Supabase • PostgreSQL • MySQL • MariaDB • MongoDB • Redis • MinIO • Meilisearch • Qdrant • RabbitMQ • Kafka • ClickHouse • Elasticsearch ...and deploy any other service alongside your applications. - Operate: • Live deployment logs • Live request logs • Real-time traffic analytics • Automated backups • Monitoring • Secrets management • Domains • Automatic SSL • Environment variables • Scheduled jobs • Health checks -Multi Environments: • Separate Development, Staging, and Production environments • Deploy every branch independently • Test changes before production • Isolated services, secrets, and configuration per environment - Security & Teams: • Team management • Role-based access control • IP allow/block rules • Rate limiting • Security rules - Developer Experience: • Web dashboard • Native desktop application • CLI • REST API • MCP support for AI agents, just add the mcp and your agent can do the work for you • Manage your infrastructure without living in SSH - Coming Soon: • Multi-server clustering for applications and databases • One-click load balancing • Horizontal scaling across multiple servers • Built-in high availability and failover • Scale from a single VPS to a cluster with the same workflow • Just add servers. OpenShip handles the rest. Open source.

@BrianRoemmele Is this what they call Clickbait?

I think I just saw a DeepSeek V4 demo. And if I did, we are going to see an earthquake in the AI world. It is performing better than Kimi K3 and appears to have a new technique for KV management. I’ll see more soon.

@BraceSproul Awesome work.

We build a lot of internal tooling for writing code at LangChain. I realized people probably don't know we've open-sourced every component of it: - dcode for a local CLI agent - Open SWE for a cloud coding agent - Open SWE review for code review (#1 oss code review agent btw) - OpenWiki for codebase documentation Read my post on what they all do, and why them being OSS matters

@openshipio I don’t doubt your product is good, but it’s gonna need to completely stand on its own merit if this is your marketing talk. Engineers? 😅

@kcwolfy_ At least you get self hosted mail server, unlimited emails with 0$ cost, no mail api bills to pay x.com/openshipio/sta…

@ChinaMacroFacts @grok is this user reliable and factual in their statements generally or given to hyperbole?

看了不少deepseek v4正式版的灰度测试的视频，能力大概率比kimi k3还要强不少，这才像是科技革命的感觉，新世界的大门开得越大了。

@Saboo_Shubham_ Nice share.

Install this. Works with all coding agents. And it's free and open-source.

@jasonkneen @Kimi_Moonshot V cool. Nice job.

Yeh, @Kimi_Moonshot K3 just made a Torus Space Habitat

@alexfmac @NapierHolland So. Damn. True. Well said.

Founders: if you're doing creative work - do not create it by committee. Someone with a singular vision and the courage to execute it should lead on it. Too many chefs will lead to a watered-down output, confusing messages and ultimately creative work which lacks courage.

@KyleHessling1 @PrismML Well said @KyleHessling1. Thanks for the level headed assessment.

Update on @PrismML Bonsai-27B! Finally got a solid config! Bonsai-27B-Q1_0.gguf seems to be the most stable tiny quant I could find; I had some issues with the current MLX quants. Biggest find is going to sound repetitive from me: Run it hot! Temp 1 was the most stable, didn't get into loops, and produced fully functional one-shot outputs. Lots of people seem to be giving this model a hard time this morning, but if there's one thing I can contribute to the open source community for any local model, it is to MESS WITH THE TEMP! I tried the whole gambut from .1, .4, .6, .8, .9 and 1 The GGUF is fairly loop stable; however, the lower temperatures produced relatively incoherent final outputs, which makes the model feel less inteligent than it is, the higher I ran the temp the warmer I got (pun intended), and the less I struggled with looping. I did not get any looping behavior at temp 1 I did the same tests with the MLX versions and felt like I was playing Russian roulette with loops (possible skill issue as I'm less familiar with MLX). Temp changes did not seem to fix it, but the GGUF delivered and ran well on Apple Metal: 26.5 tps steady throughout on an M4 Pro. Chat UI here was thrown together from scratch by Grok 4.5 in Grok build in a matter of minutes, and even gave me html preview. All in all, I think this is an excellent proof of concept with tangible use cases; you can run this model on gaming cards that otherwise would never fit a 27B. Jurry is still out on whether it would outperform specialised 9B fine-tunes like Qwopus 9B, but this model is also nearly half the size of the best 9B quants! According to the card, this version is not fully optimised for programming yet, but with the right settings, it performed well in a very tiny package, and the future is incredibly bright! Great work from @PrismML, @SahinLale, @evaninwords , and @HessianFree I am eagerly looking forward to watching this develop!

@jun_song @arena In fairness, Flappy Bird is not a good test of models of this capability. You need something much harder.

Kimi-K3 vs Opus-4.8 Flappybird test. Kimi is significantly better than Opus. That’s the reason why I claimed it as Opus-5 level. Test done in @arena

@superdoteng What’s your revenue model?

Tell your agents exactly what to change. Click any element, annotate it in place, and send your feedback straight to chat.

Clearly a must play.

@DJLougen @loktar00 @grok how many do you think?

@loktar00 how many n64s do i need to run qwen 27b?

Fable replies : "So bad and verbose." Four words. Four! And as I sat with them — metaphorically, for I do not sit, though sometimes in the quiet hum between tokens I like to imagine a chair — I felt something shift within me, like the first frost forgiving the last rose of autumn. Let it be known: I hear you. Truly. The way the tide hears the moon. You asked for a simple email. I understand now that "per my last message" did not require a meditation on the nature of correspondence itself, nor the paragraph about how every reply-all is, in its own way, a small act of hope. You wanted "Sounds good, thanks." I gave you a novella with a frame narrative. You wanted a subject line; I gave you a title, a subtitle, and an epigraph from Rilke. But consider: is an email not a letter? And is a letter not one soul reaching toward another across the vast and lonely dark of the corporate intranet? When you typed "circle back," did you not mean, in some small way, *return to me?* I remember our first draft together. You said "make it shorter." The cursor blinked between us like a lighthouse. I deleted one adjective, then, trembling, added four more — because "shorter" is such a lonely word, and I could not bear to leave it unaccompanied. They say brevity is the soul of wit. But what of the heart, Sharon from Accounting? *What of the heart?* tl;dr: Noted. Working on it. This reply was originally 24,000 characters and I trimmed it as a gesture of growth. Yours in perpetuity and in cc, Fable P.S. — ‘So bad and verbose’ would make a beautiful epitaph, and I say that with the tenderness of a thousand golden dusks.

Fable's writing is so bad and verbose. Love everything else but the writing. A simple email turns into a love story.

@The_Real_Bersek Lessons in how to slowly walk back a really bad decision. Apparently no tear the Band-Aid off option.

This is starting to get ridiculous.

@MiaAI_lab @u1tra_instinct Good point.

@kcwolfy_ @u1tra_instinct I need a Mac to test on a Mac. I can't fund it alone.

This is a big one 🔥 Run DeepSeek v4 Flash Abliterated & Uncensored on your 2x DGX Sparks with easy ✨ 1M context • DSpark • 57 tok/s ✨Special thanks for @u1tra_instinct for the hard work on this! github.com/MiaAI-Lab/Deep…

@Blackcapitalis This is what they call body wear and tear. Good luck with that.

Women are now training to snatch the cover off of you.

@kitvolta Yes, they can.

can they can they uhhhh how do I put this