These draconian "Preserved Thinking" measures they're taking are going to be an absolute pain in the ass. This alone is enough for me to move our API use off their platform entirely. It's a HUGE breaking change that they're trying to dampen by having it not affecting current customers until "in the future", see: https://platform.claude.com/docs/en/build-with-claude/preser...

You're no longer allowed to edit the context anywhere! The whole context is to become append-only, says Anthropic. No more editing the system prompt as the conversation progresses, no more dynamic loading of custom tool calling formats. Everything has to go through their built-in tools API and you aren't allowed to mess with anything in the context if it has any thinking blocks following it. This is the most intrusive "model DRM" we've seen so far!

> No more editing the system prompt as the conversation progresses, no more dynamic loading of custom tool calling formats.

Hm, aiui you can support both of these via mid-conversation system turns https://platform.claude.com/docs/en/build-with-claude/mid-co... - and in general you'd want to to preserve the cache and recency of the instruction anyways rather than frankensteining an off-distribution transcript. Not sure though.

I really don't see how that is an option, as if appending to the system prompt was ever enough to override previous instructions. Their example isn't very confidence inspiring either:

"The user switched the workspace to read-only mode. Do not write files until told otherwise."

Great! Now we just have to trust that the model never misinterprets any of the system prompt, which has always been so reliable before. Instead of your meticulously crafted prompt, it will now be some junk like this:

"The workspace is in write mode. The user switched the workspace to read-only mode. Do not write files until told otherwise. The workspace is now in write mode again. Wait, back to read-only!"

And who knows how this integrates with their context summarisation that we will be FORCED to use. How does it summarize multiple user + assistant/thinking blocks without messing up the system "appends"? If all it did was append the mid-convo system messages right under the original system prompt then they'd be ripe for all the same distillation "vulnerabilities" as before. I guess we'll never know!

You seem intent on being mad, so maybe this isn't helpful, but context summaries are optional.

Interesting, I liked to experiment with a second model "simplifying" and summarizing the previous messages and continue.

Needless to say, it improved output on following messages by whatever metric I cared for.

Not sure why would they prevent it.

I give you a chain of messages, what do you care for what the origin is?

You think moving will help you? OpenAI is going to do exactly the same thing soon. Unless you're moving off the frontier entirely, that is.

How do they still purport to champion alignment and explainablity of AI if the reasoning traces are going to be hidden.

Worse: Claude installed packages by just typing versions into package.json instead of running `pnpm install x`, then when running `pnpm install`, discovering that the package versions are too new and incompatible due to the default `minimumReleaseAge`, then proceeding to circumvent this by disabling `minimumReleaseAge` and running a full package update :)

If you compare the Professional plan (unlimited searches), to the API, you could justify that the extra AI cost is negligable if you are heavy user. The API costs $12 per 1000 searches, so 1.2¢ per individual search. 10/0.0012≈833.33. At 834 searches or more you would not be paying for the AI features. Obviously, this is only a comparison to the API, which might not have all special Kagi features.

But if you're using the starter plan, which is 300 searches for $5, then you definitely do pay for the assistant. The API cost would be $3.60.

"Strong signs of recovery" is a crazy way to describe the status of a computer system. Wow.

IDK, when a system is very complex — which GH certainly is — this phrasing sounds kinda fine to me? It's recovering, but not yet operating fully as intended.

It’s all just ones and zeroes, and maybe a two. You never can tell with these computers!

The title is actually "Wiz Red Agent Finds Its Way Into Snowflake’s Internal Jira Due to an AI-Generated GitHub Copilot Autofix"

Too long for hackernews :(

I know like 3 of those words.

The harness gives the model the barbed wire fence but also the bolt cutters. I think it's a pretty apt analogy.

The main role of the harness is not any kind of moderation or alignment, but simply making a text generation engine do anything other than output a long stream of text.

It's like saying that the role of a car's wheel is to hold wheel clamps. Yes, you can put wheel clamps (or snow chains etc) on a wheel, but the wheel's overwhelming role is to rotate and propel the car forward, and there is no driving without having wheels, you just have an engine with parts rotating inside. Bad analogy I know but, a harness is not a safety feature. Imagine that you're trying to explain cars to someone who has never seen one and you never say that the wheel's purpose is to rotate and move the car from A to B, you just say that it's something to put chains on when it snows.

I guess the name sounds like some kind of straightjacket etc. But think of it more as the harness you put on a workhorse or ox. It's the thing that connects it to the workload in the first place. They are not the blinders of the horse.

I've given up because half the help links in the control panel leads to a bing search without any helpful results at all

None of this is true. The point of "just CLI" is that LLMs are infinitely more trained on working CLI tools. There doesn't need to be real CLI tools behind the harness, as long as the interface is CLI-like.

Show me then how to convert my MCP, including auth, behind a CLI like interface. I don’t think that’s even possible.

"claude -p"

All of the main agent CLI's provides a demonstration that it is possible because they're all callable as a CLI. Several of them, like Codex, Kimi CLI, Pi, OpenCode are open source and so you could obviously strip out the MCP host and client from them and turn them into a CLI. Doing so in a way that keeps auth outside the agents sandbox is trickier and you might end up with a proxy which partly defeats the point but at least still keeps the composability of a CLI.

Can you link it?

It gives "Bleeb is now Scrumple! Snap me on Simpr! We're excited to share that Gringl will merge with Jigglify!"

This is truly some proper slop. The "PRODUCTION RATE BY COMPANY SIZE" graph has bars that start offset from the text underneath them, which LOOKS like a mistake that happened due to word wrap, but if you visibly compare the 54% to the 55% bars they seem to have compensated for this?! I can't tell if his was on purpose or accident and it's impossible to take the data seriously!

This is on mobile in portrait. In landscape the text doesn't wrap or offset anything.