Itamar Ravid

@thdxr clickhouse+hyperdx in a docker container works really well

why has no one vibed up really good local otel tooling i want to be able to look at all these traces during local dev

@trychroma Super cool, would love to try it!

Introducing Chroma Context-1, a 20B parameter search agent. > pushes the pareto frontier of agentic search > order of magnitude faster > order of magnitude cheaper > Apache 2.0, open-source

@ccccjjjjeeee zero is awesome. cloud zero now in private beta

zero vs livestore vs electricsql thoughts?

@cherkaskyb תמיכה מעולה בOTel API + MCP טובים ונוחים ממשק מעוצב טוב וארגונומי ומחיר סבבה

@iravid לא שמעתי אפילו על דאש. מה מיוחד בהם? במה הם שונים מהשאר?

אתם מקימים סטארטאפ, גייסתם 16מ דולר, וצוות הקמה של 2 תותחים, ו 4 מפתחים ״רגילים״. יש POC עובד עם כמה דיזיין פרטנרס. איזה פלטפורמת o11y אתם בוחרים לחברה החדשה שרק הקמתם?

@mitsuhiko Our setup has skills that point to mcporter invocations, credentials queried using op from 1password Works really well

No matter what I try I cannot get it to work nearly as well as skills. It feels like it was designed for agents that don’t write code, but increasingly it looks like even non coding agents will write and execute code.

The weirdest part about MCP is the actual architecture. It’s rather token inefficient for the harness but it’s also rather resource intensive for the server because it’s stateful. I think if someone were to take last year’s learnings, a new MCP would look dramatically different.

@tokengobbler ngl it does standalone app being modified would be easier to follow

@iravid feel like this always makes the demos more confusing lol

click-to-component for react, vue, svelte, astro and full devtools

@nicoalbanese10 Very timely. Love how quickly you guys are iterating

Let's go! Huge update from previous 7 day limit

@aboodman A seriously underrated aspect of LLMs. Ever since we set up Codex with skills for Dash0 and Braintrust I am so much more motivated to chase down bugs.

As amazing as LLMs are for programming, they are, if anything even MORE valuable to me for production debugging. Some people are naturally good at crawling through logs. This is not me. My eyes glaze over and I find it hard to focus. This has always been a weakness for me. Working with an LLM allows me to work at a much higher level of abstraction. I can ask the LLM to summarize the situation from the logs, then start asking questions and logically figuring out what's going on without having to try and maintain focus in a sea of text and shitty parsing scripts. I know how the system is supposed to work, but seeing how it is ACTUALLY working from log files was always very hard for me. No more. Sometimes (ok often), the AI just figures it out on its own. But often it has bad ideas and I end up getting us there. Either way, the ability to just slice and dice logs however I want with just prompting is a fucking game changer.

@tokengobbler better start hiring now

Oh my sweet summer codex

@MichaelArnaldi yeah it's errors *and* function parameters!

If you think Effect is about errors you have no idea what Effect is, it's so amazingly dumb when people propose a Result type or similar as an alternative

@kitlangton something (among many) Effect really got right is that the name is so good and amenable to combining with other nouns. the Effect Institute the Effect Platform ZIO doesn't have the same ring!

👀 Well, that was an overwhelmingly positive response. I got a lot of feedback. People mostly said shockingly kind things. But there were some great feature requests and bug reports as well. So, I just pushed an update. Ahem: - Added some prettier open graph images (see below). - Change logs are fun: effect.institute/changelog - Added volume controls - Option click to copy a deep link to any section. - The left/right seek distance depends upon the current playback speed. - Increase the brightness and consistency of the colors of the many, many borders, for some improved a11y (probably still a lot to do here). - Added some silly Easter eggs. - Fixed some of the animated code getting squished on smaller windows. - Fixed some Safari audio and animation glitches. I now hate Safari. - Typos and code-example issues. And here are some things I'm not going to change: - There will not be a text-only mode. I like this weird format and I'm going to embrace it. As evidence of both how stubbornly obsessed I am with this idea and that it's possible to get better at design, here's a prototype I made like ~5 years ago: catechism.surge.sh - There will not be a copy to markdown button. This is for human consumption. I'm not against robots, but I made effect.solutions for that. - This shall remain mobile unfriendly. There's no way I'll have enough room to do what I want on a phone. I want keyboard controls, I want space for code and text and animations. I fell in love with effect systems nine years ago, when I first learned Haskell, and worked on Scala's ZIO (Effect's direct ancestor) for years. Despite being obsessed with effects, I never would've thought they'd have a shot at becoming mainstream. I'm so, so, so incredibly grateful to @MichaelArnaldi for trying out the somehow completely non-obvious power maneuver of... simply using TypeScript. Thanks again for all the encouragement. I feel so lucky. Time to write some more chapters. 🫡

Update - Someone just shared an existing free project that does this called Vibe-Kanban. I just tried it. Super easy to get rolling. 1. Install via npx 2. Reference existing GitHub repo 3. Creating a task on the board and Claude starts working and streams output! So. slick.

@LouisKnightWebb 404

@zeeg @d4m1n Can’t go wrong with a Linea 👌👌have the mini at home with a niche zero too. Perfect setup

@d4m1n This is an ecm synchronika that I’ve had a few years. Tbqh I wish I had just bought the linea Trying to clean up the whole setup now tho

Not sure if I love it or not 🤔

@sqs I can reliably get it to hang and crash the editor by trying to find files with @

What sucks about Amp’s editor extension (VS Code/Cursor)? We get 10x more feedback on the TUI than the editor extension despite usage being roughly similar and highly overlapping. Editor folks, be picky!

@kitlangton themediterraneandish.com/spanakopita-re…

𝚜𝚙𝚊𝚗𝚘𝚙𝚝𝚒𝚌𝚘𝚗 ← 👀 TUI OTEL visualizer for Effect programs. Coming soon.

@jjenzz @benoitgrelard 8sleep is awesome tell sonny he got it right

@benoitgrelard hahaha brilliant. it’s funny with my hubby cos he is the *complete* opposite. i get a tad worried whenever he like “ooo look what i found”. he just discovered 8 sleep lol

in my old age i’ve noticed a really strong inner voice that feels manipulated by consumerism and refuses to participate. it’s basically like, “sod your ads. i’m not falling for that shit” 😅 is this common? it gets awkward when i actually need something

So basically, software engineers like to crash early and ops like to degrade gracefully. I think this exposes something kind of interesting about both disciplines and subtle about running software at scale.