sqs

I am on the Amp team. I just turned off SMS verification. Not needed anymore. Was required for ~5% of signups for anti-abuse, but I think we're good without it now.

I don't think it takes care of tone transformation (eg 他是 ni3shi4 -> ni2shi4). Or if it does, my tones are just off. But it's a really cool idea!

Using small threads (see https://ampcode.com/@sqs for some of my public threads).

If you use very long threads and treat it as a long-and-winding conversation, you will get worse results and pay a lot more.

What's super interesting is that Opus is cheaper all-in than Sonnet for many usage patterns.

Here are some early rough numbers from our own internal usage on the Amp team (avg cost $ per thread):

- Sonnet 4.5: $1.83

- Opus 4.5: $1.30 (earlier checkpoint last week was $1.55)

- Gemini 3 Pro: $1.21

Cost per token is not the right way to look at this. A bit more intelligence means mistakes (and wasted tokens) avoided.

Sorry we missed that email! I don’t know what went wrong there, but I just replied and will figure it out. This is definitely not the norm (and Build Crew is a small fraction of our users).

Yeah, I said about coding agents, “it’s obviously the future, but it’s not there yet”. That talk was from the AI Engineer conference in June 2024 (16 months ago). Coding agents have come a long way since then!

It's a big organization of teen coders who build really cool things together. Instead of coding alone, they get to hack on software and hardware projects in person and online with other smart teens all around the world.

You can see full financial and donor information at https://hackclub.com/philanthropy/ as well. Check it out. It's an organization that lots of HN folks would support (and many do). (I am on the board of Hack Club.)

Buildkite is awesome, by far the best CI product and with an amazing team.

Thank you for pointing that out. Just pushed a fix, will be live in ~5-10min.

That draft RFC you mention is superseded by https://agents.md. Now that Amp uses AGENTS.md (https://x.com/sqs/status/1957945824404729997), I made all the former agent file stuff on https://ampcode.com just redirect to https://agents.md.

Just the filename is standardized. The contents aren't, which is exactly right. From the site:

> Are there required fields?

> No. AGENTS.md is just standard Markdown. Use any headings you like; the agent simply parses the text you provide.

Amp has only been out for a couple of months, and it's growing fast. Cursor and Claude Code have been around for longer, but obviously that's not always a good thing since everything is constantly changing. Check on X, like at https://x.com/sawyerhood/status/1945865702159925734 or https://x.com/search?q=%40AmpCode&src=typed_query&f=top.

This is really cool. I used Vibe Kanban with Amp to update some of our docs and UI components, and it was great.

Priompt might be what you're looking for: https://github.com/anysphere/priompt.

Is it a very large codebase? Anything else distinctive about it? Are you often asking high-level/conceptual questions? Those are the questions that would help me understand why you might be seeing better results with RAG.

Tool-calling agents with search tools do very well at information retrieval tasks in codebases. They are slower and more expensive than good RAG (if you amortize the RAG index over many operations), but they're incredibly versatile and excel in many cases where RAG would fall down. Why do you think you need semantic indexing?

It's fundamentally hard. If you have an easy solution, you can go make a easy few billion dollars.

What a shallow, negative post. "Hype" is tautologically bad. Being negative and "above the hype" makes you sound smart, but this post adds nothing to the discussion and is just as fuzzy as the hype it criticizes.

> It is a real shame that some of the most beneficial tools ever invented, such as computers, modern databases, data centers, etc. exist in an industry that has become so obsessed with hype and trends that it resembles the fashion industry.

Would not the author have claimed at the time that those technologies were also "hype"? What consistent principle does the author use (a priori) to separate "useful facts" from "hype"?

Or, if the author would have considered those over-hyped at the time, then they should have some humility because in 10 years they may look back at AI as another one of the "most beneficial tools ever invented".

> In technology, AI is currently the new big hype. ... 10% of the AI hype is based on useful facts

The author ascribes malice to people who disagree with them about the use of AI. The author says proponents of AI are "greedy", "careless", unskilled, inexperienced, and unproductive. How does the author know that these people don't believe that AI has great utility and potential?

Don't waste your time on this article. I wish I hadn't. Go build something, or at least make thoughtful, well defined critiques of the world.

Thank you, and I’m sorry about that. Will look into this and fix on our side.

What broke on you when using Cody? Sorry to hear about that and want to fix it for you.

Congratulations and thank you to Sid, the GitLab CEO, for building an incredible company and product.

GitLab was the first code host to add more products (CI, security, ops, helpdesk, analytics, etc.) and create a whole suite, and GitHub followed. GitLab also built for the enterprise years before GitHub started to give appropriate love to the enterprise. Some people think that GitLab is a GitHub clone. Quite the opposite!

Even if you don't use GitLab yourself, you've been a huge beneficiary of the dev workflow GitLab envisioned and created, and of the competition they've given to Microsoft/GitHub. Competition in this space makes everything better.

Most users won't and shouldn't need to go through the process of adding context sources. In the enterprise, you want these to be chosen by (and pre-authed/configured by) admins, or at least not by each individual user, because that would introduce a lot of friction and inconsistency. We are still working on making that smooth, which is why we haven't been very loud about OpenCtx to end users yet.

But today we already have lots of enterprise customers building their own OpenCtx providers and/or using the `openctx.providers` global settings in Sourcegraph to configure them in the current state. OpenCtx has been quite valuable already here to our customers.

Yeah, we’re using it a lot at Sourcegraph. There are some extra APIs it offers beyond what MCP offers, such as annotations (as you can see on the homepage of https://openctx.org). We worked with Anthropic on MCP because this kind of layer benefits everyone, and we’ve already shipped interoperability.

Context is a huge part of the chat experience in Cody, and we're working hard to stay ahead there as well with things like OpenCtx (https://openctx.org) and more code context based on the code graph (defs/refs/etc.). All this competition is good for everyone. :)

Cody (https://cody.dev) will have support for the new Claude 3.5 Sonnet on all tiers (including the free tier) asap. We will reply back here when it's up.

Rule of law FTW! Governments can't usually promise timelines, but when the process is well documented and predictable, that is a very good thing.

Cody is open source: https://github.com/sourcegraph/cody. And for the reasons explained there, it makes more sense for it to be open source.

Yes, Cody can use Ollama for both chat and autocomplete. See https://sourcegraph.com/docs/cody/clients/install-vscode#sup.... This lets you use Cody fully offline, but it doesn't /prevent/ API calls to third parties; you are still able to select online models like Claude 3.5 Sonnet.

I have a WIP PR right now (like literally coding on it right now) making Cody support strict offline mode better (i.e., not even showing online models if you choose to be offline): https://github.com/sourcegraph/cody/pull/5221.

Codium or Codeium? (Edit: Original post said Codium but it sounded like Codeium to me based on the description. :)

Yeah, use this in your VS Code settings to use a different Ollama URL (here it's localhost:11434 but change apiEndpoint, model, and tokens to whatever).

  "cody.dev.models": [
    {
        "provider": "openaicompatible",
        "model": "mixtral:latest",
        "tokens": 4096,
        "apiEndpoint": "http://localhost:11434/v1/chat/completions"
    },
  ],

We should add an easier way to just change the Ollama URL from localhost, so you can see all the Ollama models listed as you can when it's available on localhost. Added to our TODO list!