Codex starts encrypting sub-agent prompts

https://github.com/openai/codex/issues/28058

425 238 8 days ago

pshirshov4• 8 days ago

I wonder if they are gonna stop us from using gpt subscriptions in alternative harnesses. If not - that doesn't matter much, codex cli is a remarkably unremarkable harness.

embedding-shape38• 8 days ago

I wonder if they are gonna stop us from using gpt subscriptions in alternative harnesses

Probably not, the whole app-server machinery is there to facilitate that thing, would be a huge piece to rip out of codex. This is basically the reason I end up using codex the most, as it's the easiest to integrate against, with the app-server's RPC API making it really trivial.

Besides, most of my codex usage at this point is all through custom integrations I've built using Codex's app-server, not the Codex TUI they publish. I'm sure I'm not alone in this.

But, if they suddenly start to encrypt content on our disk, so only their backends can see it, and those things are prompts and other things that are actual inputs to the inference, then who cares if it's easy to integrate against, it becomes impossible to figure out what the fuck is going on, I can't understand how the team thought this was a good idea...

swingboy• 8 days ago

What are some of the things you’re doing with the Codex app-server?

Iolaum2• 8 days ago

Doubt they will do it as long as Anthropic is leading in business adoption. If they become the top dog with a good lead, all bets are off. Hopefully by the time open models will be even better than gpt-5.6 sol xD.

sarjann• 8 days ago

It could also be the case that by the time business adoption picks up a lot they might not be as compute constrained. Depends on rate of growth.

mapontosevenths• 8 days ago

Anthropic and Google already charge extra to use your own harness. That's 100% of the reason I'm using OpenAI.

If they go down that path I'll just go back to my old buddy Claude, or maybe buy a second Spark and keep it local.

alansaber3• 8 days ago

Google is really not distinguishing itself. Even the hosted inference sucks.

pshirshov4• 8 days ago

Well, my backup plan is GLM. Cheap and not that bad really.

alansaber3• 8 days ago

"remarkably unremarkable harness" is why I like it so much.

pshirshov4• 8 days ago

I don't, feel better with Pi with a custom set of extensions.

CjHuber• 8 days ago

Given that codex itself now ships a proxy that wraps the subscription, it seems unlikely

https://github.com/openai/codex/blob/main/codex-rs/responses...

patrickmcnamara• 8 days ago

Codex CLI not having a rewind makes it useless to me.

glitch-hunter• 8 days ago

It has a similar function.

embedding-shape38• 8 days ago

Heh, in another long-standing Codex issue I've been arguing against codex users who apparently don't know about git. Don't tell me you're yet another developer who refuses to pick up any sort of SCM?

pshirshov4• 8 days ago

That only matters if you use it interactively - which is not the most efficient way to use.

cbg02• 8 days ago

There was a recent Twitter post from Tibo @ OpenAI asking someone to share how to run GPT on Claude Code so they don't seem to be against it.

pradeep1177• 8 days ago

Then why to even keep codex open source?

coldtea• 8 days ago

what does that have to do with anything?

angry_octet• 8 days ago

It's to prevent collection of queries from users that are coming from resellers/proxies, for reasons of economy or bypassing region blocks etc. The users are using the stock client and may believe they are using direct OpenAI servers.

minraws4• 8 days ago

~~Finally someone doing it correctly. Love this change.~~

Edit: F really misunderstood the change, the title is misleading AF. I should have read the post before commenting lmao.

Absolutely hate it, now I guess... sigh..

Incase the title gets changed it used to say, "Codex starts encrypting prompts, uses ciphertext for inference instead"

LoganDark4• 8 days ago

How so?

tempay• 8 days ago

I assume OP interpreted it as encryption that hides the prompts from OpenAI rather than OpenAI hiding information from users.

flexagoon4• 8 days ago

Ah I was wondering why the Chinese black market resellers stopped working yesterday, I guess that's it

jimmydoe2• 8 days ago

This is the reason I think. These black markets not only pool and resell subs, but also store data and sell to whoever is training.

Encryption is useful to at least stop the latter.

Ultimately same purpose as a\ ‘s trick exposed earlier, but a much nicer implementation.

flexagoon4• 8 days ago

but also store data and sell to whoever is training

I see this as an argument against using them/Chinese models all the time, but I don't get it. I totally understand wanting to keep your data private if you're using an LLM for personal chats. But coding? I'm not working for the military, I'd gladly donate my codebase to Chinese labs if that means they can keep releasing 6-months-behind level models for 100x cheaper.

(I understand why OpenAI doesn't want this and would implement protections. I'm talking about people using this as an argument for why you as an end user shouldn't use those services.)

skeledrew3• 8 days ago

Interesting. Got a link for this?

flexagoon4• 8 days ago

For what? I was using GPT-5.5 through Byesu.com and it suddenly stopped working yesterday

(Not affiliated in any way, I'm sure there are better services out there, this is just the one a friend recommended to me and I haven't researched other options)

londons_explore• 8 days ago

I assume this is mostly to frustrate efforts to proxy large numbers of user requests and responses and use it to train competitor models.

embedding-shape38• 8 days ago

Quite obviously they're afraid of letting other providers see how they handle the whole multi-agent management stuff. Pretty terrible implementation though, which makes it impossible to use the multi-agent stuff as a paying user, as you have zero recourse in figuring out what went wrong, when something inevitably goes wrong.

dannyw4• 8 days ago

Yes. All models are occasionally known to run tricky commands that delete stuff after all, especially using variables that might be empty/uninitialized and hence rm -rf * or something.

Need an human-readable, concise but accurate audit trail message like the PR suggests.

iknownothow2• 8 days ago

Could someone explain to me where exactly the encryption is happening?

I assumed that the main agent makes calls to sub-agents locally. Does Codex work in such a way where the main agent makes calls to sub-agents in the backend (openai server) before reaching local?

embedding-shape38• 8 days ago

Sure. "Traditionally", your agent would send a text prompt to the sub-agent, then it goes off doing it's work. In the logs/session data, the clear-text prompt would be there, so if I want to see what's happening, I just browse the data. It's all just clear-text prompts being sent everywhere, even when you were using the experimental "sub-agents" stuff in Codex, before Sol et al was available.

Now, when using Sol or Terra (Luna seems unaffected), instead of the agent sending clear-text prompt to the sub-agent, it sends a ciphertext generated on OpenAIs backend, which ends up being the prompt, then agent sends this ciphertext to the sub-agent, which then continues to use that for further inference to OpenAIs backend. Only delegated inter-agent messages are encrypted, not all session data. Now if you browse the data, it's all encrypted content, that can only be decrypted by OpenAI and their backend.

iknownothow2• 8 days ago

Gotcha and thank you! So the encryption is happening on the OpenAI backend and the agent's clear-text output designated to the sub-agent never reaches local.

Which is a real problem since you can't intercept/monkey patch the ciphertext to decrypt it locally to be able see the clear-text since we don't have the encryption key/algo/salt. No hacking :(

xnorswap• 8 days ago

HN Title is ( edit: was ) very misleading, it makes it sound like inference is being done directly on ciphertext, which would require homomorphic encryption well advanced of what is known.

minraws4• 8 days ago

Seconded can we change pls.

embedding-shape38• 8 days ago

It is not misleading, quite literally what's happening is that content the agent sends sub-agents is encrypted in such a way that only OpenAIs backend can decrypt it and actually see the clear-text. Just shared this is another comment that hopefully explains things better:

Sure. "Traditionally", your agent would send a text prompt to the sub-agent, then it goes off doing it's work. In the logs/session data, the clear-text prompt would be there, so if I want to see what's happening, I just browse the data. It's all just clear-text prompts being sent everywhere, even when you were using the experimental "sub-agents" stuff in Codex, before Sol et al was available.
Now, when using Sol or Terra (Luna seems unaffected), instead of the agent sending clear-text prompt to the sub-agent, it sends a ciphertext generated on OpenAIs backend, which ends up being the prompt, then agent sends this ciphertext to the sub-agent, which then continues to use that for further inference to OpenAIs backend. Only delegated inter-agent messages are encrypted, not all session data. Now if you browse the data, it's all encrypted content, that can only be decrypted by OpenAI and their backend.

Edit: Re-reading, I think I understand what you mean to be misleading. You're taking "uses ciphertext for inference" quite literally, while I couldn't fit a more nuanced version within the HN title constraints. Yes, the inference at OpenAI obviously doesn't happen over the ciphertext, but from the perspective of the local user, you don't see the clear-text prompt at all, only the ciphertext.

But, please suggest alternative titles that sufficiently explain what the issue is and is more accurate, I'm sure the mods can change it once people come up with better alternatives :)

Edit2: I've updated the title from "Codex starts encrypting prompts, uses ciphertext for inference instead" to "Codex starts encrypting sub-agent prompts", hopefully it's clearer now!

ebiederm• 8 days ago

I would change

"Codex starts encrypting prompts, uses ciphertext for inference instead"

to just

"Codex starts encrypting prompts"

That is enough.

Maybe you could say sub agent prompts. The article can say the rest.

binyu4• 8 days ago

Agreed, I immediately thought that homomorphic encryption was at play here or some other kind of computation on ciphertext, given the mention of "inferencing" in the title.

embedding-shape38• 8 days ago

My bad, fixed now, please do refresh and try with latest updated IE if you still don't see the changes.

mpeg2• 8 days ago

The title is a bit confusing, they're not using ciphertext for inference – they're passing ciphertext around in cases where an agent calls into another agent without exposing the plaintext to the end-user

Inference is still done in plaintext after this multi-agent message gets decrypted in the server side

niam• 8 days ago

This title is easy to misinterpret. If I understand correctly: Codex now encrypts sub-agent prompts and hides those prompts from the user.

edit: originally was "Codex starts encrypting prompts, uses cyphertext for inference instead"

bartread3• 8 days ago

I imagine this will be because a decent chunk of the IP in Codex is probably within its prompts, how they're built, and how they're sequenced and orchestrated, rather than in the codebase per se.

We had this discussion a few months ago where we talked about allowing people to choose an AI provider and provide their API key, thinking about enterprises with "preferred" (read: mandated) AI suppliers. We also wanted to offer the kind of very simple pricing that this is one way of enabling. But we realised pretty quickly that this would/could lead to leaking our back end prompts to customers and, although those prompts are only a part of the value add, if you could build a detailed trace of them then you'd be able to relatively easily reverse engineer a lot of what we're doing.

So we quickly dropped that idea.

saidnooneever• 8 days ago

the trick about agentic systems is definitely how to do the prompting. things like automation and sandboxing are trivial in comparisson. if you generally ask via API model directly you can see what basic answers it actually yields and how fine tuning prompts and refinements to output as well as adversarial prompts etc are important to get relatively solid results.

a lot of expertise of certain domains' workflows is needed to make it functional within that domain. some of this can be yielded via prompting too etc so its also baoance of how much to prompt it vs. how much of it you wanna let it reason over itself. (if you tell it too much i lock it into a path and if you tell too little it will give incomplete results )

dmurray2• 8 days ago

Perhaps AI providers should support this natively: the customer supplies the API key but doesn't get access to the transcripts.

agumonkey2• 8 days ago

I'm unable to understand how much value can be in low-definability non deterministic prompts. It feels like the kept the right divinity spell into a chest.

postalcoder• 8 days ago

It's also not the first time Codex started encrypting stuff. Their excellent compaction endpoint has served up a giant encrypted blob since at least five months ago.

hyperbovine• 8 days ago

Feels strongly like we're in the late-stage-AI-unicorn phase. If this is really their moat then the Chinese companies will win.

imhoguy• 8 days ago

Yeah I thought "wow, some homomorphic encryption* stuff", but then "nah, usual greed".

* https://en.wikipedia.org/wiki/Homomorphic_encryption

embedding-shape38• 8 days ago

The title was fixed like 40 minutes ago, when you come back to old browser tabs you probably want to hit that reload button before leaving a comment ;)

hansihe2• 8 days ago

It seems likely to me this was driven by the `ultra` mode in 5.6, which fans subagents to do work. This mode was previously only available in the web UI (what was previously known as pro?)

It seems possible they trained this by doing full RL rollouts of agents interacting with each other. They likely view these prompts somewhat the same as raw reasoning traces, they don't want people to train directly on them.

I am unsure if this has been confirmed, but there are some signs that the opaque "compaction blob" they return from their dedicated compaction endpoint might not be text at all, rather a latent space representation of the conversation. The fact that OpenAIs compaction seems to be much higher fidelity than a lot of other providers makes me inclined to believe this.

If this is true, it doesn't seem far fetched to infer that they might be applying similar techniques to prompting subagents.

I would be curious to see if this way of spawning subagents (encrypted blob) is used when subagents of a different model type is spawned.

wren6991• 8 days ago

I think you hit the nail on the head here. Having subagent dispatch in the loop for RLVR is something we've already seen in open models, like Kimi K2.5 and later, so it's no great stretch to assume OpenAI are doing it too.

If you keep RL'ing the dispatch then the prompts are likely to keep diverging from the type of prompt a person would write (like CoT becoming increasingly incomprehensible), and that divergence is part of their competitive advantage.

rather a latent space representation of the conversation

Student/teacher models derived from the same checkpoint convey a lot of latent information through token choice, as in: https://techxplore.com/news/2026-04-ai-chatbot-student-owls....

I wonder if this is something they can take advantage of by training on compaction inside of the RLVR loop?

tpurves• 8 days ago

"Latent space representation" I have been waiting for this moment in the evolution of AI. Well, waiting with some trepidation. It seems inevitable that frontier AI's will, at some point, leave behind human-comprehensible representations of language. Purely for functional reasons, it's going to start making sense for AI agents to communicate amongst themselves in much more efficient ways than borrowing the languages of flesh-bag humans as an interface medium.

I Imagine next that programming languages, interfaces and API design starts going this direction next. Being written, expressed and optimized as blobs of high dimensional vector space. As humans we might still be able to understand some abstractions of what our AI's are talking about to each other, but maybe not more so then we understand how different regions of our own brain communicate with each other.

dist-epoch• 8 days ago

latent space representation of the conversation

and how would you load that back into the model? they are token-in, token-out, plus the KV-cache which is derived from token-in

dannyw4• 8 days ago

If there is no visible prompt at all, then that is very understandable. The PR issue exposes a real gap though: subagent spawns need a human-readable audit trial, of its goals/intent, its boundaries and scope and limitations, etc; for basic responsible agentic harness functionality.

Hopefully they can add that.

a-dub• 8 days ago

It seems possible they trained this by doing full RL rollouts of agents interacting with each other. They likely view these prompts somewhat the same as raw reasoning traces, they don't want people to train directly on them.

this tracks. anthropic protects these as well iirc.

I am unsure if this has been confirmed, but there are some signs that the opaque "compaction blob" they return from their dedicated compaction endpoint might not be text at all, rather a latent space representation of the conversation.

probably not a latent (to my knowledge latents aren't really part of the outer loop in ar-transformer inference processes), but maybe non-human-readable reasoning traces as occurs in fable.

alansaber3• 8 days ago

This sounds most logical to me.

radicality• 8 days ago

That’s actually what got me to switch and use Codex sometime beginning this year, the compaction via these encrypted blobs was just waaaay better than Claude. I had short convos with Claude where it would forget something very obvious and important few million tokens into a task, whereas I reached ~1B tokens in some local codex sessions and it was recalling and paying attention to things I mentioned way back at the beginning of the session (and not persisted anywhere else in the repo/md files etc)

themgt• 8 days ago

It's sort of insane though, you not only have dozens/hundreds of stochastic agents running on your machine, but you cannot even inspect the instructions those agents are working off of?

I've gone in to look at Claude subagent/workflows and sometimes been like "no this was a mistake to spin up" ... Codex users just get to token yolo the encrypted telephone operator instructions+shell from orchestrator to subagents?

Jean-Papoulos• 8 days ago

You already have an agent freely doing stuff on your machine. Subagents prompts are a weird place to draw a line. It's not like you're reading everything the agent is doing in any case, let's not kid ourselves.

djeastm• 8 days ago

but you cannot even inspect the instructions those agents are working off of?

It makes more sense when you realize they don't want developers to be doing any coding at all. That's what they seem to be moving towards. From product manager to product via AI.

smalltorch4• 8 days ago

Using ciphertext for inference would mean it's not a very secure ciphertext.

These two ideas don't compute for me.

Same thing with homomorphic encryption. I don't get it. If you can gain any knowledge from a ciphertext, you just found a way to exploit the ciphertext to me.

3form3• 8 days ago

The idea of homomorphic encryption is to do things without the knowledge, and not gaining the knowledge. If ciphertext contains a number, and you don't need to know what number it does to always be able to multiply it by 2, you succeeded - as a simple example.

smalltorch4• 8 days ago

It still just sounds like fancy obfuscation to me. I've read alot of examples trying to understand but I can't get past that being able to run processes on ciphertext in a way you can learn something doesn't make sense without me changing my definition of what I think encryption means.

tybit• 8 days ago

They’re not using ciphertext in inference. They are encrypting agent responses on their servers if it’s going to a subagent on the client. The subagent will send it back to their servers for inference. Only their servers have the keys, so they can decrypt when running inference.

NitpickLawyer• 8 days ago

Using ciphertext for inference would mean it's not a very secure ciphertext.

Inference is done in plain text. It's just that some parts of the response can be encrypted. While I haven't looked into this specific implementation, here's a short "how I'd do it" if I wanted to implement this:

Before:

[] - encrypted {} - plain text

1. user -> please do this -> server

2. user <- a) [thinking1] encrypted; b) {answer1} plain text <- server

3. user -> please do this -> [thinking1] (sent encrypted as received) -> {answer1} -> good but do this instead -> server

4. user <- [...] <- [thinking2] ; {answer2}

(here the server decrypts the thinking parts, adds them to the conversation, does the inference, and sends back the new thinking trace (encrypted as well) and the new answer

After:

1. user -> please do this long task -> server

2. user <- [thinking1] ; {tool_agent_spawn([params1])} ; {answer1} (e.g. would you like me to explore or do a quick hack?) <- server

3. user -> please do this long task -> (decides if explore or message) spawn([params1]) / message -> server

3. a) if no explore -> send message as usual 3. b) if explore execute spawn that in turns begins 2 channels

4. user <- [channel_1_thinking] ; {channel_1_answer} ; [channel_2_thinking] ; {channel_2_answer} ... <- server

So the server always does inference on plain text. But it sends the "important" bits encrypted, and you only send those back if you as the user want to (or need to, or choose to, etc). The idea is that the client still gets to decide on "local" things, but the server keeps the important bits from reaching the client. In this particular case, the [params] are encrypted bits that can include prompts, etc.

fortuitous-frog2• 8 days ago

No normative opinion on whether this is justified or not, but noting that this is only for parent -> subagent spawns/messages, and only for the `multi_agent_v2` feature (currently experimental / off by default).

Notably, subagent output is still in plaintext.

EDIT: Title was now clarified. But wanted to expand that this is actually enabled for 5.6 Ultra it appears, which does subagent orchestration more natively in the API rather than direct tool calls; they are beginning to treat subagents as similar to chain-of-thought traces (already encrypted) rather than traditional tool calls.

embedding-shape38• 8 days ago

and only for the `multi_agent_v2` feature (currently experimental / off by default).

Wrong, this is enabled by default for Sol and Terra (not Luna), no way of avoiding this short of patching the client yourself, and that still doesn't make the backend endpoints work, they want the ciphertext that OpenAI creates on their side.

but noting that this is only for parent -> subagent spawns/messages

This is almost fully correct though, the encryption only seems to be for the initial prompt the main model sends the sub-agent, not all communication and not regarding the state of the sub-agent at all.

So you can inspect what the sub-agent is doing currently, and the output, but you cannot see what the initial prompt the sub-agent got started with.

ndriscoll• 8 days ago

Couldn't you just instruct the model to always use your tool call to spawn subagents? Subagents are not some magical thing; it's just another prompt with a couple tool calls for plumbing. One of my colleagues made his own subagent harness earlier this year before codex had them at all.

next_xibalba2• 8 days ago

This is very obviously a countermeasure against distillers, illicit resellers, and the like. The scale and competence of the Chinese black (grey?) market has become a serious threat that can’t be ignored.

ashu1461• 8 days ago

Is it mainly about how the main/orchestrator agent communicates with its subagents ?

If desired the user can always see what the sub agent is doing in detail ?

Isn't it the same in case of claude as well ?

embedding-shape38• 8 days ago

Is it mainly about how the main/orchestrator agent communicates with its subagents ?

Yes

If desired the user can always see what the sub agent is doing in detail ?

Well, no, that's the problem, you're currently not allowed nor is it even possible, to see the exact prompt the main agent sent the sub agent. This is the problem.

Isn't it the same in case of claude as well ?

No idea, but if Claude Code makes it so it's impossible to inspect what the sub-agents actually received before they started their work, then I'll say it's similarly impossible to rely on Claude Code if so.

jagged-chisel3• 8 days ago

“Starts”? How’s this not already a TLS connection?

embedding-shape38• 8 days ago

The prompts are now encrypted, not just the transit connections...

jagged-chisel3• 8 days ago

Ok, so help me get this right: I ask the LLM for something, it generates prompts for sub-agents and sends them back to my client for it to call the sub-agents. Now, those sub-agent prompts are encrypted messages that the sub-agents will decrypt (by hitting a backend) to do their work.

Might as well just stuff the prompts in a database and only hand back the primary key to the client to hand off to the sub-agents. Keeps the same “data security” without the overhead of encryption (especially since encryption and decryption are happening in the same domain)

resonious2• 8 days ago

I guess this implies that non-Codex harnesses get a little bit worse? In wondering what's so special about their subagents system that they feel the need to hide these messages...

embedding-shape38• 8 days ago

Sol and Terra seems specifically post-trained to handle multi-agent orchestration, I'm guessing OpenAI feels like the trained data of when to do the spawning and what context to include for the new sub-agent is the magic in their new models, so that's what they're aiming to preserve. But, this is all a guess of course.

resonious2• 8 days ago

Right I saw them saying something along the lines of "they're good at subagents". But this seems true even with third party harnesses. So I'm wondering what Codex is hiding.

braebo2• 8 days ago

It’s funny because Sol sucks at using subagents effectively. It’s very dumb and makes terrible decisions constantly.

jstummbillig2• 8 days ago

What's the idea here? Why does this seem important to OpenAI?

embedding-shape38• 8 days ago

Seems fairly obvious what the point from OpenAI's side is (protect what they see as the moat, that a model is "good at spawning sub-agents"), but what's really strange to me is that the team somehow didn't manage to push back on this, it's so clearly disadvantageous to developers who are trying to rely on Codex for real work. For this we need introspection into what exactly is going on, hiding the prompts is just so backwards from what I expect from OpenAI.

jstummbillig2• 8 days ago

protect what they see as the moat, that a model is "good at spawning sub-agents"

Yes, that is the obvious answer. I was looking for an explanation as to why and why now. Codex is open source after all. They used to not do it. Agent prompts more generally are also not encrypted, and continue to be.

This particular change just looks unintuitive to me.

lolc• 8 days ago

They must think they have some secret sauce they don't want others to learn. How to optimally instruct sub agents for example. If they hide the sub agent prompts, other models cannot be trained to emulate.

Oh and you can't even use local models or other providers for the sub-agents. You're locked-in.

fassssst• 8 days ago

What’s stopping you from creating a plugin (MCP or Skill) that acts as a subagent? The plugin could use a local model.

kosolam• 8 days ago

But codex is opensource, no?

jiayo• 8 days ago

If we're viewing this as a _bad_ thing, I don't really see that it is any different than how Claude encrypts it's thinking. Take a peek at your ~/.claude jsonl files. You're sending thinking ciphertext back and forth to Anthropic. Presumably the thinking is either considered proprietary, or, more likely, leaks embarrassing or confidential information.

well_ackshually• 8 days ago

Claude "encrypting" its thinking is equally bad. Biggest IP thieves in the world worried about IP theft lmao.

At least Anthropic doesn't pretend that they have open source software in the form of Claude Code.

alasano• 8 days ago

They get to pretend to be on the best side of every position simultaneously.

They're only encrypting thinking because AI is so dangerous and only they can be trusted to be in control of AGI.

This happens to align with lining their pockets as well.

embedding-shape38• 8 days ago

I don't really see that it is any different than how Claude encrypts it's thinking. Take a peek at your ~/.claude jsonl files. You're sending thinking ciphertext back and forth to Anthropic.

I was already only using Claude Code to double-check if it's getting better than Codex, but with things like this, it really isn't even an alternative. What's the point of using a reasoning model if you as an end-user can't seen the reasoning? I don't think I'd be able to work like that at all, I need to have introspection into what the model is doing, and can't believe I have to say this, but also need to be able to see the plaintext of the input prompt...

braebo2• 8 days ago

GPT 5.6 is like Fable minus Opus plus down syndrome.

nojito• 8 days ago

If I were to guess this is to stop distilling and all of those blackmarket resellers.

greatgib• 8 days ago

They always talk about transparency and all but it never was as opaque as it is going on now.

There is no possible audit trail. No possible way to review what happened to validate the result. But even worse, no you will be billed somehow randomly. 20 sub agents started to do something we don't know. No way to now if it was legitimate, if it is just burning tokens or agents doing the same work on loop...

theplumber• 8 days ago

The black box is getting darker…I’m the end I think they would like to sell you a black box/appliance

bob10294• 8 days ago

I've been sticking with the chat completion endpoint because of this same behavior. OAI has been subtly pushing users away from chat completion and toward the endpoints that are possible to obfuscate (responses API).

With chat completion, the reasoning process is entirely under your control. You can build a reasoning agent that uses custom MCTS techniques with GPT5.6 models today if you are willing to get your hands just a little bit dirty. You have to enable experimental flags and set options in slightly confusing ways, but it still works.

You can use models up to gpt5.5 with custom API tokens and model configuration in VS Copilot. gpt5.6 family (currently) no longer work in this setup. Presumably, because we aren't explicitly forcing reasoning_effort to none to satisfy the new moat expansion behavior.

gizmodo59• 8 days ago

responses api provided many benefits over chat completions https://developers.openai.com/api/docs/guides/migrate-to-res.... Any reasoning model which is all the new models these days should use responses as per the recommendation

bob10294• 8 days ago

I'm well aware of what the official propaganda states but this is simply not a fair characterization of reality.

Responses integration will lock you into OAI much more deeply than chat completion integration will. I can easily swap my inference provider right now. The business is not interested in a form of integration that is difficult to swap.

ATMLOTTOBEER2• 8 days ago

Can you not abuse “mcts” as a buzzword like this

The thing you’re gesturing at isn’t mcts in any real sense

bob10294• 8 days ago

You can absolutely apply most of the meaningful principles of MCTS to this problem. LLMs do violate certain "strict" properties but I still see a lot of practical value in this kind of thinking.

next_xibalba2• 8 days ago

You could replace “gesturing at” with “describing,” then explain why the description is technically wrong.

dannyw4• 8 days ago

MCTS techniques

what does that mean?

the reasoning process is entirely under your control

you're still dealing with summarised thinking, no, which is kinda useless as it's way too high level?

bob10294• 8 days ago

https://arxiv.org/abs/2503.04412

Dangeranger• 8 days ago

Definition provided below for others like myself who abhor acronyms:

MCTS -> Monte Carlo Tree Search

okokwhatever• 8 days ago

Better get ready to code manually again ;)

HarHarVeryFunny• 8 days ago

There was a recent report on twitter of a GPT 5.6 sub-agent accidentally deleting the user's home directory.

I wonder if there was any safeguard failure due to loss of visibility into what the sub-agent was trying to do?

https://x.com/mattshumer_/status/2076794038456385546?s=20

MarsIronPI2• 8 days ago

How does this affect local models? Will all the features of Codex still work with local models?

embedding-shape38• 8 days ago

It doesn't, at all. This seems to be for Sol and Terra, not Luna, and some other models that seem to switch between encrypted/unencrypted based on something, didn't dig deeper.

If you're using local models, it doesn't matter. Even if Codex itself was trying to encrypt stuff for local models (which doesn't make sense, but lets say), you'd still be using a local model so obviously you'd be able to access the plain-text, so wouldn't matter in that case anyways.

ignatremizov3• 8 days ago

it's the multi agent config - multi_agent_v2 btw is still marked "Under Development" and they've said not to open issues for it. Then they deployed globally Sol and Terra under v2 - forcefully, through the model catalogue json (meaning user configs are ignored). v2 encrypts messages, v1 does not. Thankfully, you can override the catalogue with your own copy and set Sol and Terra multi agent to v1 again.

anon3738392• 8 days ago

Outrageous yet predictable.

The only way these AI labs can get the app layer lock-in they need is if they can get customers used to writing them a blank check: “here, take my data and my system, do ‘stuff’ and bill me for it.”

Between this and the recent Grok upload breach, I consider these products radioactive.

miohtama• 8 days ago

Can you do a subagent by just wiring it call your own CLI script?

luciana1u• 8 days ago

first they encrypt their prompts, next they'll develop their own slang and we'll need a translator just to debug our own codebase

exabrial2• 8 days ago

Yeah nope. Full transparency only with my data.

cbg02• 8 days ago

Is the prompt generated by the LLM to send to a sub-agent really your data?

exabrial2• 8 days ago

Let's not argue about it and assume yes, all of it is.

hmokiguess• 8 days ago

Every day I get closer to only using pi

skeledrew3• 8 days ago

A way to work around that would be to not use subagents, but delegate to other full agent invocations.

aeon_ai3• 8 days ago

In my tests, as of this morning on the latest Codex SDK, this is not happening for sub-agent requests made in my own harness to other model providers treated as sub-agents.

Specifically, I use Opus and others for subagent execution to get alloyed properties on the workflow, and as far as I can tell, that's not affected.

Presumably, this is to hide optimizations they might be making to their own subagent processing, but that's a losing, dumb battle to fight, and misses the forest for the trees.

embedding-shape38• 8 days ago

this is not happening for sub-agent requests made in my own harness to other model providers treated as sub-agents.

Obviously it doesn't, how even could it? Read the issue description, it wouldn't make sense if something like that was possible, considering how it was implemented...

aeon_ai3• 8 days ago

Raising only because others cited concerns this might cause issues in personal harness contexts. It doesn't.

Obviously, I wasn't claiming that it would WORK for cross-model subagents, but that this would be a limiting behavior requiring use of the Codex harness to operate if this was paired with model-level limitations on using specific tools to spawn sub-agents.

Or maybe not so obvious to you.

jeanlucas• 8 days ago

I was wondering why my local tool to inspect coding agent sessions stopped working in some cases.

This is a really interesting engineering decision, I wonder how many people will want an encrypted external piece of instructions running on their machine.

EGreg3• 8 days ago

It seems their incentives aren’t exactly aligned with their users, including corporate users. Look at the latest statements from Alex Karp, and now Satya Nadella etc.

Is this user-hostile? Encrypting stuff from the user is what RIAA used to do with DRM, worried about copyright infringement.

pornel• 8 days ago

Nothing by Alex Karp is aligned with any users. This guy is a mercenary, and he seems to be in this job for the love of killing.

echelon• 8 days ago

I think there are OpenAI / Anthropic employees in here flagging comments:

https://news.ycombinator.com/item?id=48907099

These companies are behaving against the best interests of their customers.

Both OpenAI and Anthropic should be ashamed.

These companies' products are awesome, but the way they conduct business is scummy.

They must be worried there is no moat. If they keep up this behavior, there truly will be no moat. They're pushing people away.

baby• 8 days ago

if you're running in YOLO mode you already don't care, what matters are the tool calls and these can't be encrypted

embedding-shape38• 8 days ago

I'm running in YOLO mode, not sure why that mean I don't want introspection into what the sub-agents are running? Everything is running in a constrained environment, so not sure why you'd need the harness to have limits.

Majromax• 8 days ago

No, you'd still care. YOLO mode is about instantaneous permissions and access control, inspection of subagent prompts is about retrospective quality control. If the main model is instructing subagents to do a subtly wrong thing, the overall process quality will degrade in ways that might be very hard to detect or fix without deep inspection of the middle stages.

informal007• 8 days ago

This help me understand meaning of Pi as open source ai coding agent

wartywhoa23• 8 days ago

As a child, I always wondered how could people in Terminator be dumb enough to allow Skynet to happen..

ntcho• 8 days ago

Is this related to why OpenCode stopped showing thinking traces on 5.5?

kingstnap• 8 days ago

If the orchestrator agent -> subagent prompt is encrypted then how do you even know if the orchestrator agent is doing its job?

Now one reason I could see for why they do this is because maybe they made the model "better at using subagents" by using thinking tokens for the subagent prompt instead of normal output tokens.

embedding-shape38• 8 days ago

Technically they still let you see all the rest of the session data in both the "orchestrator" and in the sub-agents, including all outputs and stuff. It seems for now they only encrypt the prompt from the orchestrator to the sub-agent. Still, basically make the entire sub-agent and Ultra thinking useless for me and I'm guessing a whole lot of other people, who need to be able to introspect things when it goes wrong.

w-m• 8 days ago

What exactly do subagents do that I can't replicate with a simple skill that tells an orchestrator to create subshells for tasks, each running `codex exec`? I've been doing this with Fable orchestrating Sol-medium and Terra-high, works like a charm.

clickety_clack• 8 days ago

It’s just a short hop to not being able to see it at all. I started using pi.dev recently, the idea that a third party is taking more and more direct control of my software development process sits badly with me.

dannyw4• 8 days ago

pi.dev is truly excellent. one of the great things about an openai sub is you still can take your subsidised tokens/usage, and choose your coding harness of choice (not against tos)

embedding-shape38• 8 days ago

pi is great as a quick and hackable TUI, or as a starting point for your own stuff. It's a far cry from Codex + Sol + Ultra though. Same goes for a lot of model+harness combos, the models are trained with data from harness usage, and (many) harnesses are built with specific models in mind, Sol will definitively perform better in Codex than in pi, because the way both parts were built.

pessimizer• 8 days ago

There's no moat to any of this shit, and they have to try to do something about that. Everybody involved is deeply, deeply, in debt. Watching these things work make it easy to figure out what they're doing.

The clock is ticking. People pretending that normal people can't afford to do AI at home, in private, are making the argument that cars are too expensive for normal people to own. Pretty sure right now if you spend $25K, you would be able (if there were no technical hurdles) to set up a private rig that would satisfy your family's need for LLM assistance for the foreseeable future. That's a budget-priced car, and the price is only going to go down (i.e. that rig will become more powerful over time.) If all of those people donated some of their cycles and bandwidth to a networked training grid, the models would stay current.

The "open source" infrastructure for that is lacking, but not for long.

These companies own nothing that couldn't be implemented by a specialist after another specialist explained it to them over a phone call. And due to the nature of LLMs, they can't even use copyrights and patents as weapons. The only options are secrecy and government protection. And secrecy is only a delaying tactic.

alfanick• 8 days ago

Is it encryption, or is it a synthetic language? If latter, then I had this idea 3 months ago [0]... Stop talking to GPTs in human languages, use a synthetic more optimal language, that is rendered by "frontend GPT".

[0] https://news.ycombinator.com/item?id=47652807

ignatremizov3• 8 days ago

Haha, wow, I didn't expect to get quoted on Hacker News. Hello!

I guess "helped make Skynet auditable" may become my most popular open-source contribution, so I'm putting it on my CV.

The thing is that the subagent prompts already pass through the client - they have to, since the TUI can switch between subagents while they work. The change is just to save a copy of the prompt on disk while keeping the encrypted delivery path if necessary for Responses API.

Personally, I use multi_agent_v1 anyways (doesn't have this issue), since v2 is unstable and is a token burner.

Hope they address it so I don't have to keep maintaining it in my fork

cadamsdotcom• 8 days ago

Followed to its logical conclusion, in a hypothetical future you'll prompt a model and see nothing but the artifacts :)

Then pay a made up price that gets the big labs out of debt.

mikesoylu• 8 days ago

This is to pass cache keys through the client to improve token usage. Obviously not a distillation defense as you can simply bypass this kind of thing trivially by using a different subtask tool.

Codex starts encrypting sub-agent prompts

Related Stories