Annoying and alarming things about OpenCode

Conclusion Stop using OpenCode.
Post-script: Local LLMs This is worth its own post – I have multiple attempts in my blog drafts – but it needs to be addressed briefly here. My opinion on local LLMs like Qwen3.6-27B is they are corrosive to the stability and conceptual fidelity of your codebase in the same way as frontier models, with the following three differences:
You avoid the uncanny valley where the model appears to be intelligent before doing something stupid; the stupidity is self-evident and this helps calibrate your interactions.
The weight count is too low to reproduce the training set verbatim, which nudges the calculus on whether the output should be considered tainted. This is distinct from larger models which can reproduce inputs verbatim, but are trained to refuse to.
You avoid supporting or relying upon cloud providers.
I’ve had useful results from input-oriented tasks like: “I think there is a bug in code x with symptoms y, my guess on the mechanism is z. Read all relevant code, come back with a call chain and code citations.” Framing it as a search problem reins in the clanker’s propensity to make shit up.
Using LLMs for code generation feels like a dead end. However thoroughly you think you understand your architecture, your planning is constantly undone by shortcuts like “what if I just move this mutable state into the middle of the design so everyone can share it?” This is hostile to your ability to understand your code, beyond the fact that you didn’t write it.
Drawing answers directly from knowledge in model weights leads to hallucination even for multi-trillion-parameter models, so why bother making them that big? If people were realistic about limitations then we wouldn’t be building new power stations for datacenters, and they wouldn’t be rammed into every product.
The entire software ecosystem around LLMs is completely rotten, and if they do ever become “just a tool” then some actual systems engineering needs to be done around them to turn them into tools instead of security black holes. That work will have to be done by humans.

t0mpr1c33• 2 days ago

> Don't use Opencode, Don't use remote models (cloud providers), Don't use Docker to isolate coding agents.
May as well write a post saying don't use LLM's for any SWE work.

Alternatives exist

kristopolous6• 2 days ago

Opencode offers free models with 200/requests over 5 hours. That's why I use it. It is the only reason I care about.

What's the alternative there? Gemini used to have free tiers. Qwen used to. AMP used to. Ollama cloud used to. Codebuff used to... None of them have those programs anymore

The reason I use it is purely financial. I do not have an employer and I'm writing free software

Edit: pi.dev doesn't have free inference endpoints. That's the constraint.

ernsheong• 2 days ago

pi.dev

calumcl• 2 days ago

You could easily just use Pi or even a Codex fork to avoid Opencode as a harness and still pay zero dollars.

kristopolous6• 2 days ago

Can't get those generous free quotas from opencode with that.

rafram2• 2 days ago

There are other OSS agent harnesses.

kristopolous6• 2 days ago

That offer free inference?

I've written my own. I know they exist.

I don't offer free inference on large models however

dizhn4• 2 days ago

Currently only Mimo has a usable free offer with their own CLI. (opencode clone)

chuckadams2• 2 days ago

If you don’t know what OpenCode is, imagine a boot stamping on a human face forever. The boot is made of TypeScript and the face is everything we have learned about security and systems software since the invention of the electronic computer in the 1940s.

I nominate this for a Bulwer-Lytton prize in the Tortured Metaphor category.

wren699111• 2 days ago

If you don't know what OpenCode is, imagine a dark and stormy night.

jauntywundrkind• 2 days ago

Referring to the top comment on yesterday's submission for Terranc McKenna psychadelic bad trip. https://news.ycombinator.com/item?id=48968515

jareklupinski• 2 days ago

perhaps if the boot were Rust it would crumble in the face of Bare Metal

evnp• 2 days ago

"There will be no curiosity, no enjoyment of the process of life. All competing pleasures will be destroyed. But always— do not forget this, Winston— always there will be the intoxication of power, constantly increasing and constantly growing subtler. Always, at every moment, there will be the thrill of victory, the sensation of trampling on an enemy who is helpless. If you want a picture of the future, imagine a boot stamping on a human face— forever."

George Orwell, 1984

saghm• 2 days ago

That would be an example of a non-torturted metaphor. When you copy the famous quote verbatim and then substitute both parts of it for things that are almost entirely unrelated, it becomes tortured.

Gander5739• 1 day ago

Smells of AI writing - just look at those emdashes.

xg152• 2 days ago

As tortured as you will feel after using OpenCode!

volf_• 2 days ago

I switched from OpenCode to Pi and there was a big improvement in terms of tool calling performance and I find the experience less buggy.

OpenCode has also seemed to have disappeared from https://openrouter.ai/apps/category/coding

awinter-py3• 2 days ago

yeah I noticed it missing on the openrouter leaderboard -- very interested to know what that's about

dv_dt• 2 days ago

probably about their own model routing service zen?

nthypes• 2 days ago

How you deal with LSP? One of the best things in OpenCode is LSP integration.

638d127673• 2 days ago

I haven't tried it myself but there seems to be an extension for this: https://pi.dev/packages/pi-lsp

the_mitsuhiko2• 2 days ago

How you deal with LSP? One of the best things in OpenCode is LSP integration.

I would love to learn how LSP helps you. I did various basic evals with LSP and I did not find it to improve the agent performance at all. Maybe on newer models this gets better because they are now RLing with LSP in the context, but at least in the past having a commit hook that runs lints and typecheck at the end, is more token efficient than having every edit be injected extra LSP results. You are almost guaranteed to be wasting tokens since partial edits are unlikely going to result in type checks passing.

(Disclaimer: I work on Pi)

lkjdsklf2• 2 days ago

I use a forked version of the pi-lsp. It works pretty well out of the box, but I just tend to fork extensions by default and then hack them up how i see fit.

resonious• 2 days ago

Oh My Pi has it. I'm a big OMP shill right now. Seems not very popular, but it has the stability of Pi with the features Opencode (and more I think; OMP has web browsing and a more advanced edit system too). OMP often outperforms Claude Code and Codex for me.

melodyogonna• 2 days ago

I disliked OpenCode's LSP integration, as Pyright kept distracting the model from what I wanted done in a Django project. I prefer to run MyPy at my own convenience

jcul• 2 days ago

I tried opencode and pi recently.

Having come from Claude code, I was a bit shocked that both seem to allow edits by default with no confirmation dialog.

One IIRC allows you to enable it in config and the other needs a plugin.

InsideOutSanta3• 2 days ago

Having come from Claude code, I was a bit shocked that both seem to allow edits by default with no confirmation dialog.

There are confirmation dialogs in OpenCode. You can configure them using the permission config, same as Claude Code. Having said that, as a general rule, none of these harnesses are safe to run on your local file system.

shiggity• 2 days ago

OpenCode requested to be removed from OpenRouter rankings: https://github.com/anomalyco/opencode/issues/11926#issuecomm...

kristianp• 2 days ago

Why is stated:

a lot of apps run 100% of traffic through openrouter while for us it's a small subset
but companies we were trying to work with used openrouter as a source of truth on popularity so they thought products that were much smaller were bigger than us

ffamac• 2 days ago

I purged OpenCode when I noticed it downloads npm packages in the background without asking the users first. This exposes higher risk of supply chain attach risk.

BeetleB4• 2 days ago

I switched from OpenCode to Pi and there was a big improvement in terms of tool calling performance and I find the experience less buggy.

That's not because OpenCode is bad, but because Pi is good. I can say the same about Claude vs pi.

fallingbananna• 2 days ago

Kilo is the top 3 option there, and that's a fork of OpenCode.

tomaytotomato• 2 days ago

My bet:

In a years time the discussions on using CC, Opencode, Pidev etc. will be redundant as we will be building our own tooling with whatever programming language and tooling you want.

Some will repurpose a popular opensource tool or harness, others will build it from scratch in an hour or so.

blkhawk• 2 days ago

unless someone fixes the issues with LLMs at a fundamental level those will all have the same kinds of issues as is described with OpenCode and they will all be horrendous.

LaurensBER• 2 days ago

I feel that a better title for this article would be: "Some minor annoyances that, when fixed, would improve OpenCode"

# Prompt Cache Misses

It globs your filesystem and re-reads AGENTS.md (injected in turn-0 system prompt) on every SSE turn. If you put a quick note in AGENTS.md to be read in the next session, you immediately force a full re-evaluation.
Personal favourite: it puts the current date in the turn-0 system prompt and re-evaluates every SSE turn. If you’re using OpenCode at midnight you get a full prompt cache miss.

Okay, I can live with those.

# Compaction

Want to sit for 10 minutes while the LLM server prefills the entire session with a new prompt prefixed to it, just to turn it into 5 bullet points that go at the top of a new session? Me neither. I get what they are going for, but I’ve not seen it work well. Neither compaction nor pruning is implemented well, and they interact poorly.

Is this an OpenCode specific issue? I've seen the same with Codex and Claude

# System Prompts

The default system prompt is opinionated (fine) but it has shit opinions (not fine). It took me a while to figure out why my agent kept saying “Use ABSOLUTELY NO COMMENTS” when dispatching subagents.

Okay, so change it? Any LLM is opinionated, this system prompt enforces consistency across different models which seems reasonable.

nextaccountic4• 2 days ago

I feel that a better title for this article would be: "Some minor annoyances that, when fixed, would improve OpenCode"

Those are labeled as "annoying things"

Read also the "alarming things"

cpburns2009• 2 days ago

> > The default system prompt is opinionated (fine) but it has shit opinions (not fine). It took me a while to figure out why my agent kept saying “Use ABSOLUTELY NO COMMENTS” when dispatching subagents.

So that's why OpenCode likes to randomly remove my comments.

dinfinity• 2 days ago

Opinionated system prompts forcing this are awful. It's similar to putting "ALWAYS INDENT WITH TABS" in the system prompt, but worse, because "self-documenting code" is a convenient lie lazy developers like to propagate.

I have the following in my instructions, but I often need to remind agents of it because they follow the shitty system prompt instructions:

"ALWAYS include MANY inline code comments describing what blocks of code are supposed to be doing. Inline comments serve as inline specification, a parity check between the code and the specification, and are a means to _communicate_ with all future programmers, including yourself. Write Once, Read Many. The code needs to talk to whomever is looking at it in natural language."

cevn2• 2 days ago

I actually share that opinion of no comment from LLMs. I just dont want to read them. So as usual, an opinion is… an opinion.

naasking• 2 days ago

Comments sometimes help the LLM think through a problem. They are particularly useful if you have thinking disabled, eg. it's not uncommon to run Qwen3.6 locally, but its thinking is extremely verbose, so it's fast if you disable thinking, but doing that will sometimes cause it to think a little in comments, and if you disable that too, then it will produce worse output.

Comments can always be stripped reliably later.

lynndotpy• 2 days ago

This article has several sections and you are commenting on things under the section titled "Annoying Things".

There is another section titled "Alarming Things", with a subsection titled "It’s Fucking Full of RCEs". (Those are RCEs other than those which simply derive from the issues pointed out in previous subsections).

brianjking• 2 days ago

Definitely agree on the title. Also could have worked as 'Some minor annoyances that, when fixed, would improve OpenCode. Also, I don't like AI"

zyuiop2• 2 days ago

Did both of you intentionally skip the part of the article about the critical security issues?

Hizonner• 2 days ago

> The default system prompt is opinionated (fine) but it has shit opinions (not fine). It took me a while to figure out why my agent kept saying “Use ABSOLUTELY NO COMMENTS” when dispatching subagents.
Okay, so change it? Any LLM is opinionated, this system prompt enforces consistency across different models which seems reasonable.

OpenCode embeds scatters bad coding policy all over all of its inbuilt prompts, including tool instructions. The only way I've found to fix them is to edit the source code. And, yes, they're also huge and repetitive.

A framework's native prompts should include only information about how to interact with the framework itself.

verdverm2• 2 days ago

yet OpenCode is #1 harness per Artificial Analysis

https://artificialanalysis.ai/agents/coding-agents?coding-ag...

disiplus2• 2 days ago

Cache Misses are pretty bad, i have a locally running deepseek v4 flash, i have tuned it now to have 1100-1300 prefill. Its not great but properly useable. Imagine having a session with already 100k and half of it has to be prefilled it would be waiting minutes with worse numbers. And if you are paying by the token, for hosted models, you are wasting money.

nylonstrung3• 2 days ago

The bigger point is that the codebase is hopelessly bloated by vibe coded features at this point, mirroring the same problems with Claude Code. And stability, performance, memory use have all gotten terrible.

I liked Opencode a lot previously but there's no doubt it's not well-written software. Pi has fully replaced it for me and there are quite a few newer options that learned from Opencode and exercise more minimalism where it matters

alightsoul2• 2 days ago

Can you name a few? I'm looking for alternatives

snapplebobapple2• 2 days ago

Got a link? Pi is suitably as to make search engines questionable

dofm2• 2 days ago

I am on this trajectory too; opencode got me started but I tried to understand the way it is working and to do that I would normally read the source code, but the source code is really not fun.

JeremyNT• 2 days ago

I keep seeing these conceptual complaints about OpenCode but it's unreasonably effective, so it's difficult to justify any time spent in changing to something else.

It hardly matters that it's vibe coded and ugly if it works well... and it does (for me).

Saving a few CPU cycles and reducing RAM utilization on my local system isn't going to meaningfully improve anything in and of itself; the wall time is all bounded by the model or by the code it's invoking (i.e. compiling or running the test suite).

versteegen• 2 days ago

I always run opencode with envvar `OPENCODE_DISABLE_AUTOCOMPACT=1` because I discover a new horrible bug in its autocompact every time I don't... including the config .json option to turn off autocompact not working.

Sad but not surprised to hear about all these other issues. I like opencode, it does plenty right but plenty wrong.

xg152• 2 days ago

I feel that a better title for this article would be: "Some minor annoyances that, when fixed, would improve OpenCode"
If you don’t know what OpenCode is, imagine a boot stamping on a human face forever. The boot is made of TypeScript and the face is everything we have learned about security and systems software since the invention of the electronic computer in the 1940s.

Regardless if that's a good or bad take, I'd like titles on HN to reflect what the author wanted to say and not editorialized to what HN readers think the article should be saying.

intothemild2• 2 days ago

That's the thing. You can't change it. They do not accept PRs from what I've seen. Whilst it's open source it's only in name. There may be the odd PR accepted from outside the team. But very very rarely.

It's honestly embarrassing to call this open source.

Pi on the other hand, whilst they have their own unique issues with PRs, if you came to them with a genuine problem, or fix, they seem to accept it.

Now I point back to those issues you pointed out, all have PRs... Unmerged. Same with the infinite depth subagent calling.

wasabi991011• 2 days ago

Open source doesn't mean that it has to accept PRs. It has never meant that the code author's must accept community contributors.

OpenCode has an MIT license, which allows you to copy/fork the code and modify it how you see fit.

Please review the open source initiative's definition of open source: https://opensource.org/osd

k-langton3• 2 days ago

Hey! I work on OpenCode, so thank you for this measured reply ;)

Just wanted to add a few things:

- We actually don't do tool-call pruning any more. But compaction is a necessary evil for the time being. There's only so much context window, and if you wish to keep working on the same task for a long time, the model needs to summarize its current progress so it can continue without hitting its token ceiling.

- With V2 (in beta atm), we have a new system for handling changing system instructions (AGENTS.md files, available skills, etc.) that avoids cache misses at all costs, while still giving the robot the newest information. I recorded a little video about it here:

https://x.com/kitlangton/status/2075749116760457346/video/1

vitamark2• 2 days ago

Can you shed some light on what's the deal with the new interface (web/desktop)?

I updated one of my opencode containers today and had to downgrade it back because:

1. Worktree/workspace support is gone altogether. No way for agent to work in parallel on different things safely except prompting it directly to create a worktree.

2. Seeing multiple sessions properly now requires opening tabs for those?

3. In general, switching between several projects/sessions/whatever stopped working, and command palette is hotkey-only.

Issues about the new design just sit in an endless unsolved pile on GitHub which seems hopeless. And an agressive stale bot is not helping with that.

I came to OpenCode because it seemed hackable and simple, and at this point it seems like a "move fast break a lot" project that you got to live with.

(rant over, sorry if this is addressed to the wrong person, just needed to vent a bit)

steve_adams_86• 2 days ago

Does V2 finally imminentize the eschaton? V1 has completely failed at this so far

sgallant• 2 days ago

Just checked out the video...you, my friend, could be a professional voice actor (assuming that job still exists in the future).

mrighele• 2 days ago

# Prompt Cache Misses

It globs your filesystem and re-reads AGENTS.md (injected in turn-0 system prompt) on every SSE turn. If you put a quick note in AGENTS.md to be read in the next session, you immediately force a full re-evaluation.
Personal favourite: it puts the current date in the turn-0 system prompt and re-evaluates every SSE turn. If you’re using OpenCode at midnight you get a full prompt cache miss.

I used opencode with deepseek v4 via openrouter during the weekend. Openrouter claims a cache hit rate of about 95%. Maybe it can be improved, but it doesn't seem broken.

I agree that it is bloated, and uses a lot of memory and cpu. But I don't remember seeing any bug during its usage, or at least not big enough to notice

trollbridge2• 2 days ago

DeepSeek has a phenomenal cacheing architecture that specifically can work with less-than-ideal clients like this and get cache hits anyway.

whythismatters• 2 days ago

When I looked into opencode a few months ago I was struck by the default system prompt which came across as messy and random, iirc it also contained a typo or two. It made me look elsewhere and I didn't look back.

tomComb• 2 days ago

IMO, if you "put a quick note in AGENTS.md" then you should expect a cache miss.

I think that is reasonable and intuitive.

That was the first complaint I came across in the article and it made me question the level of indignation of the whole thing.

jmkni• 2 days ago

posted just as I start to use OpenCode haha

petesergeant2• 2 days ago

You should absolutely be running your AI agent inside _some_ kind of sandbox. I put together a list of 19 mostly open-source ones here: https://pleasedonotescape.com/ along with a list of non-project-solutions

wren699111• 2 days ago

Yeah, I don't think that line quite landed. My point was that LLMs are inherently adversarial (read, "relentlessly proactive") and sandboxing should be a first-class integral feature of your harness, not an afterthought. The problem with dev containers is you still tend to end up with something you care about on the same side of the trust boundary as the LLM.

petesergeant2• 2 days ago

LLMs are inherently adversarial (read, "relentlessly proactive") [and] you still tend to end up with something you care about on the same side of the trust boundary as the LLM

I mean this is a problem with many coworkers too, so you deal with it in the same way: limit what they can do to creating pull requests.

cowpig• 2 days ago

`greywall -- opencode`

singpolyma3• 2 days ago

This is an anti AI post masquerading as an anti opencode port.

Nevertheless I'm curious what people might suggest to use instead with local models.

nextaccountic4• 2 days ago

Maybe https://github.com/tontinton/maki

lucideer11• 2 days ago

- Sensible permission system - when the agent runs git diff && rm -rf /, what do you think will happen in your current coding agent? It will treat it as git *. Maki uses tree-sitter to parse the bash command and figure out the permissions requested are git * and rm *. Disable using --yolo.

This sounds identical to what OpenCode does (&, contrary to their own marketing, does not seem sensible at all).

The fact that this mechanism is so primitive across all agentic clis is frankly a serious indictment of the competence & experience behind LLM-focused development teams.

This is a security issue that has in the past been solved via explicit file-path-based allowlisting, & more recently - due to file-path matching being deemed insufficiently secure - apps have been switching to cryptographically signed binary matching as a more modern alternative to the less secure path matching. In that context, the fact these recently-written "modern" agentic clis are using a method that's more basic & even less secure than the "old insecure" method most apps use is pretty shocking.

lucideer11• 2 days ago

I agree generally but I'd moderate this to say this is an "anti- the current state of AI" post insofar as none of the issues listed are inherent to LLMs, they just seem to be ubiquitous in the current generation of agentic CLIs. I'm not aware of any CLI that doesn't have the listed severe issues, but I could certainly write a CLI myself (if I had the time & energy) that avoided these issues.

archargelod• 2 days ago

Good sentiment, wrong target. I took some liberties to fix the beginning:

If you don’t know what AI is, imagine a boot stamping on a human face forever. The boot is made of shit and the face is everything we have learned about security and systems software since the invention of the electronic computer in the 1940s.

codelion• 2 days ago

Finally good seeing someone else also getting frustrated with OpenCode and posting the issues.

alanwreath2• 2 days ago

I’m not really here to defend opencode (or bash it, who has the time?), but in setting up SDD paired with the right tools to both make deterministic black boxes out of certain actions, paired with subagents. I have ran with opencode and qwen 3.2 on a 5090 and I’m getting results on a production codebase of golang where a review of changes from Claude had no changes but acceptance of the solution.

I may be missing some aspect of where the local llm’s are losing, but I’m content with draining my local tokens and treating the frontier models like a less junior SWE.

drbscl• 2 days ago

Admittedly I prefer Hermes and oh-my-pi, but most of this post is just hyperbole

urvader• 2 days ago

About the caching things. This is a flaw in the backend, not in the Frontend. If the backend is only relying on prefix cache you need to look for more fine grained cache solutions like HiCache or Lmcache. If your backend discards the whole cache because of one date is changed, you really can’t blame it on the Frontend. It is an implementation detail that you push to the wrong side.

wren699111• 2 days ago

Prefix caching is inherent in how global softmax attention works. Modifying the prefix invalidates the suffix because the suffix's cached KV projections are a function of the hidden state after the model has evaluated the prefix.

There have been some attempts to approximately stitch KV blocks without full re-evaluation (see: CacheBlend on arxiv), but the results aren't promising.

You mentioned HiCache: as far as I can tell this is a prefix tree cache that efficiently shares prefixes across related histories. If you change the prefix, everything after the change still misses the cache and needs to be prefilled again.

NitpickLawyer• 2 days ago

That's not how it works. While there are some research papers looking into using chunks of kv cache while discarding others, those involve architecture changes in the models themselves. HiCache / LMcache (used in sglang/vllm) are ways of optimising the juggling of kv cache, but IIUC it still needs to be token perfect atm.

maxloh• 2 days ago

Their "OpenCode" naming is controversial too.

Quoting moozilla's comment 11 months ago (https://news.ycombinator.com/item?id=44741894):

---

If anyone is curious on the context: https://x.com/thdxr/status/1933561254481666466 https://x.com/meowgorithm/status/1933593074820891062 https://www.youtube.com/watch?v=qCJBbVJ_wP0

Gemini summary of the above:

- Kujtim Hoxha creates a project named TermAI using open-source libraries from the company Charm.

- Two other developers, Dax (a well-known internet personality and developer) and Adam (a developer and co-founder of Chef, known for his work on open-source and developer tools), join the project.

- They rebrand it to OpenCode, with Dax buying the domain and both heavily promoting it and improving the UI/UX.

- The project rapidly gains popularity and GitHub stars, largely due to Dax and Adam's influence and contributions.

- Charm, the company behind the original libraries, offers Kujtim a full-time role to continue working on the project, effectively acqui-hiring him.

- Kujtim accepts the offer. As the original owner of the GitHub repository, he moves the project and its stars to Charm's organization. Dax and Adam object, not wanting the community project to be owned by a VC-backed company.

- Allegations surface that Charm rewrote git history to remove Dax's commits, banned Adam from the repo, and deleted comments that were critical of the move.

- Dax and Adam, who own the opencode.ai domain and claim ownership of the brand they created, fork the original repo and launch their own version under the OpenCode name.

- For a time, two competing projects named OpenCode exist, causing significant community confusion.

- Following the public backlash, Charm eventually renames its version to Crush, ceding the OpenCode name to the project now maintained by Dax and Adam.

efficax• 2 days ago

Wrong adam. The adam who cofounded chef has nothing to do with opencode

tesnorindian3• 2 days ago

speedping• 2 days ago

If OpenCode is so terrible it wouldn't be the base for many of the fortune 500's internal cli coding agents, Meta included

Supermancho• 2 days ago

If OpenCode is so terrible it wouldn't be the base

Ignoring the appeal to authority inherent to this kind of statement, it's ignoring another implication. Large companies can afford to maintain a fork. Using the latest release of OpenCode day to day is a different proposition than serving as some corporate developmental branch "base", as shorthand for all sorts of changes/usage.

t0mpr1c33• 2 days ago

I'm not sure that I buy into the premise that everything Meta does is OK

pelagicAustral• 2 days ago

My conclusion is that OpenCode is clown-car turboslop with a security posture of “let me bend over for you daddy”

My type of writing.

persedes• 2 days ago

Gonzo style / polemic would make a good filter on here.

alfiedotwtf• 2 days ago

Weird this is currently #1 on the front page.

mistic92• 2 days ago

I have moved to kilo code for now

k-langton3• 2 days ago

that... is an opencode fork

LoganDark• 2 days ago

Cache misses are the one reason I stopped using OpenCode in favor of Pi. OpenCode mutates the system prompt every turn, which is completely unacceptable and betrays such a basic misunderstanding of LLMs that I can't in good conscience trust the rest of the product regardless of how technically impressive it is.

progx• 2 days ago

I've had the same experience. Tried it and use in 2 hours more token than in a whole week. Normally i got 80 - 90 % cache hit rate.

acd• 2 days ago

Use Picode as an alternative to Opencode

rs_rs_rs_rs_rs• 2 days ago

With posts like this there's usually priors involved, it's crazy vitriolic.

swsieber• 2 days ago

Textual command filtering is entirely useless. It is fit for no purpose. Nobody with any instinct or experience in security would even bother to implement this filter because it achieves nothing except a false sense of security.

Not if you only use an allow list. E.g. only allow things matching a particular prefix to run. Also, isn't this what every agent does? And by every I mean Codex because I haven't used the others.

lucideer11• 2 days ago

Good to see Codex as a datapoint: I haven't used it. I can confirm though that this is certainly what Claude does, with the addition that Claude's adherence to its own string-match based allowlist seems even less deterministic than OpenCode's.

mring33621• 2 days ago

There are definitely annoyances with OpenCode.

The remote-preferred attitude caught me once. Also, its server APIs don't match its docs, if you want to talk to it programatically.

The following config works for me to keep it on a local model:

{ "$schema": "https://opencode.ai/config.json", "provider": { "local-llama": { "npm": "@ai-sdk/openai-compatible", "name": "Local llama.cpp Server", "options": { "baseURL": "http://localhost:8080/v1" }, "models": { "Chefs Choice": { "name": "ANY" } } }, ...

This allows me to put any model I want on port 8080.

qarl218• 2 days ago

It globs your filesystem and re-reads AGENTS.md (injected in turn-0 system prompt) on every SSE turn. If you put a quick note in AGENTS.md to be read in the next session, you immediately force a full re-evaluation.

Um. Which then gets immediately cached again.

I prefer this behavior. I prefer my changes incorporated immediately, rather than wait for the next session. Call me crazy.

(Author is unhinged and this story is upvoted because #AI-HATE.)

wilkystyle• 2 days ago

Also, aren't instructions earlier in the prompt treated with more weight than those that occur much later? Sure, you could insert the system prompt later to avoid re-caching as much, but then the system prompt (if I am understanding correctly) carries MUCH less weight.

(And also your point about immediately being re-cached a good one)

qarl218• 2 days ago

Yes, exactly. This whole complaint boils down to "I don't want my changes to take effect when I make them because I don't expect that behavior."

sokoloff• 2 days ago

I also would want anything that I saved into AGENTS.md to be incorporated right after I saved it.

I might accept if my newly saved information was overlooked, but it's hard for me to fathom how I'd demand "I want to save this information in AGENTS.md now, but I don't want you to look at it until the next time I launch you..." and call any behavior other than that annoying or alarming.

mmaunder• 2 days ago

Re local LLMs: “You avoid the uncanny valley where the model appears to be intelligent before doing something stupid; the stupidity is self-evident and this helps calibrate your interactions.”

Making the authors stance on LLMs clear.

t0mpr1c33• 2 days ago

It's a tool. Use it like a tool. It's not a substitute brain.

recursive• 2 days ago

Clarity is a hallmark of good writing.

axegon_• 2 days ago

There are a number of points I completely agree with in this blog post. Security implications, handing out all your data to anthropic/openai/google, allowing a slop machine to execute arbitrary code on your machine and having full access to your network is something that would have made anyone working in security commit a ritual suicide just thinking about it as recent as 2022-2023. And I am baffled by the fact that we all sign tons of NDA's as part of our employment contracts, just to throw all out the window by giving it all away to anthropic/openai/google etc, at what is effectively a symbolic fee. We all know it costs a lot more than 100 bucks a month or whatever it is they charge.

That said, I firmly believe that if AI is to survive, the future HAS TO BE local or near-local. Having said that, statements such as

Docker causes security holes:
It creates a god-service that runs as root.
It deliberately punches a hole in ufw firewalls.

Sorry, none of those are correct IF you know what you are doing. Though I will admit, seeing people that know what they are doing is increasingly uncommon.

Also there is nothing wrong with developing inside containers. If anything, that is arguably one of the biggest selling points for containers - environment(s) you can crash infinitely at no cost.

I still dislike opencode for a bunch of reasons - the assumption that llms are immune to screw ups, being one. As for the default behavior - using cloud by default - I didn't know that(I do not use any AI for direct coding tasks) and if that is the case, yeah, this is bad. Undeniably a horrible decision.

wren699111• 2 days ago

Yeah, this is one of the weaker parts of the post because I didn't really make my position clear: that sandboxing should be a first-class integral part of every harness, not an afterthought that risks putting something you care about on the same side of the trust boundary as the LLM.

Some of my own frustrations with bad Docker deployments leaked through. I did snip some of them out way back when I posted this, but maybe not enough.

Joker_vD• 2 days ago

that sandboxing should be a first-class integral part of every harness,

Essentially, what you want is that the people who produce harnesses like OpenCode to also implement proper sandboxing and security. Why... why do you believe they'll do anything remotely satisfactory?

It's been empirically demonstrated, again and again, that at least 85% of software developers (and this is a very generous estimation) can not be trusted to bother about the security, so leaving the implementation of the security mechanisms and policies up to them simply doesn't work in practice. It just doesn't.

lucideer11• 2 days ago

This is a good summary of the dangers of using agentic clis, but the title & general focus on opencode is odd for two reasons:

1. Most obviously & importantly this is a complaint without a straightforward suggested alternative. A sibling commenter mentions suggesting fixes to Opencode would be more productive: I don't necessarily agree since many of these issues are fundamental & would likely require an almost ground up rethink & rewrite, but the issue is that the article contains no constructive proposal at all: it may as well be titled "Stop Using LLMs"

2. None of the major issues listed are unique to OpenCode. At least the full list within the "Alarming Things" seems in my mind to apply to Claude CLI, & I would guess most other agents from frontier model providers.

Granted it's worthwhile documenting these issues as a plea for someone to build better tooling from the ground up, so the article is far from worthless - on the contrary I've bookmarked it & will be sharing & referencing it widely & often. But the title & focus is just very odd & seems misguided, especially when the contents of the piece is otherwise so good.

qarl218• 2 days ago

I wonder how he imagines you can give shell access but securely prevent arbitrary command execution.

His complaint that "echo git | bash" still works is particularly nutty.

wren699111• 2 days ago

Sure, the two main approaches are revoking execute permission from the executable itself, and write-protecting the things that it would modify (like $PWD/.git). Both of these are achievable with sandbox-exec or landlock_restrict_self(2).

This works for the `bash` example too. Spawning a new shell and running arbitrary commands is not an escalation because the child shell inherits the parent's restrictions.

lucideer11• 2 days ago

This is very doable - it's a shell you're creating within the cli so you have a lot of control over the environment. If you look at tools like sandbox-exec & apparmor, they specifically target binaries, not string input. E.g. if I write a bash script & put it in my path, OpenCode's allowlists aren't looking at the commands my bash script is exec-ing, nor monitoring subshells it creates. A proper sandbox has full context of the execution chain. Not only that, it's potentially much more powerful & expressive in that you can allowlist indirect filesystem access from commands run.

For example, I recently ran opencode via sandbox-exec & naively granted the sandbox access to the aws cli. First attempt to exec it failed because I hadn't granted the sandbox access to the credential file in ~/.aws. This took 2 seconds to resolve but effectively highlights how utterly useless the tree-sitter string-parsing they're doing is. It's so concerning because implementing tree-sitter for that is non-trivial to the point that any competent dev should surely realise quickly while implementing it that this avenue is utterly useless. Yet every high-profile agent cli has taken this approach.

TZubiri• 2 days ago

I'm glad that when I recommended opencode to a friend, I taught them first to install a vm and create an app specific user to boot.

Many users did things right the first couple of days and then turned the security off after seeing it 'work fine' on its own.

Was a cool thing to use for a couple of weeks in v1, you have to be quite static to keep using it at v68 of self vibecoding while the llm providers like openai already developed their in house alternative.

asveikau• 2 days ago

My opinion on local LLMs [...] :
You avoid the uncanny valley where the model appears to be intelligent before doing something stupid; the stupidity is self-evident and this helps calibrate your interactions.

This seems to be capturing a feeling I am finding messing with local LLMs: popular software around this seems to be slop coded by no talent hacks. It's a common experience to see a python process pegged at 100% CPU with a huge amount of memory allocated, seemingly doing something that could be a simple shell script. That is before stuff reaches the LLM.

If that runs on somebody else's machine, you don't notice. When you see it in top it's frustrating.

thomasjb• 2 days ago

It's not a great UI, but for messing with ideas in a VM, the free tier is very hard to beat, otherwise I'd be on Claude Code, but still on a VM, because I won't let them near anything I actually care about.

nirinor2• 2 days ago

I use CC at a client because that's their stack. I use (a particular version of) OpenCode for my own work, because it is so much better. Reading this post makes me sad.

Because everything matches and explains oddities I've seen and forgiven. So despite the hyperbole and the places where the authors' general sentiment are not mine, he's basically right, and I need to find a different harness.

If the post resonates for you too and you've found something better, happy for recommendations. pi seems to get the most mentions here, anyone feel it (or another option) is specifically better on these architectural aspects?

ethbr18• 1 day ago

I'd offer that the AI harness ecosystem is discovering the same thing any rapidly-growing software space did before it -- lines of code are a maintenance burden.

That's not something a team sees when they're doing greenfield, forward development. (Most projects, now)

It only crops up when they have enough on-the-books code that features / bugfixes start requiring introspection of existing code. (Now to ~1 year from now)

And a lot of these projects don't seem to be approaching codebase size growth from a future code debt perspective...

nirinor2• 5 hours ago

Answering myself: I've tried pi and oh-my-pi (omp.sh) in the last couple of days, and the latter feels solid and more effective than CC/opencode/pi. I have not reviewed the code, and 2 days is not a lot, so take with a large grain of salt, obviously.

drdexebtjl• 2 days ago

Regardless of its flaws, OpenCode is the harness I’ve been the most productive with by far, and I’ve tried them all.

These are all just some minor annoyances, some disagreements, and most importantly, a fundamental misunderstanding of the point of command filtering. It is not for security, it’s for steering.

Anyway, it didn’t sound like the author tried building anything with OpenCode, and if they did, they said nothing about the most important part: how well did it do?

mlnj• 2 days ago

Regardless of its flaws, OpenCode is the harness I’ve been the most productive with by far, and I’ve tried them all.

I feel the same way. It's the right amount of 'gets out of my way' and 'won't nuke my machine'.

Especially the ease of using Plan mode and quickly getting stuff done.

mutkach• 2 days ago

Any recommendations for non-typescript-based coding TUI harnesses/agents other than Codex? Given that Codex is open source someone should've built an untethered client?

nylonstrung3• 2 days ago

2 of the most promising Rust ones

https://github.com/1jehuang/jcode https://github.com/gi-dellav/zerostack

673dfddnd• 2 days ago

I just did the obvious thing and asked an AI to plan / consider and implement improvements / changes in the code based on the article.

intothemild2• 2 days ago

No mistakes

bitwize• 2 days ago

Well, out of the agentic harnesses available, OpenCode is the most promising and capable whilst still remaining open; proprietary alternatives like Claude Code have all these problems and more. So I guess the only way out is to just not use LLMs for development at all.

...

[chadjak.png] Your terms are acceptable.

ltbarcly3• 2 days ago

Opencode's cache hit rate is better than claude code. Opencode has in my experience been better than any of the closed source / closed development alternatives and is not locked down.

https://news.ycombinator.com/item?id=48883275

nylonstrung3• 2 days ago

Have you tried Pi though? I was a big Opencode user but once I started using Pi I never looked back

rattlesnakedave• 2 days ago

Pi, especially bundled as OhMyPi, performs significantly better than anything else.

Zababa• 2 days ago

The mix of humanizing the LLM, calling it "clanker" and being very aggressive towards it is really weird. I don't think it's a good habit to take, it feels like it could bleed into how you interact with people. Many interactions are through text interfaces these days.

all2• 2 days ago

The mouth speaks what is in the heart. What is inside eventually leaks out. In anonymized spaces, this is especially true, and has been acknowledged from the beginnings of the internet (and perhaps long before). When you don a mask, your heart speaks freely.

But the knife cuts both ways. As you speak, so will your heart be. I think modern psychology supports this. Actions drive attitude, and attitude drives actions. If you assert your will in either place, the other is disrupted and will adjust.

You can literally change yourself into whoever you want to be. But a passive participant in this process will simply show the world what kind of person they are without ever realizing it.

jacobgold• 2 days ago

I really hope truly open models and local inference are the future.

But are there are any reasons a pragmatic and informed developer would use OpenCode vs Claude/Codex as a harness today?

I'm not seeing anything competitive in terms of cost/quality/trust?

simonw3• 2 days ago

Including the date in the system prompt - at the cost of a cache invalidation at midnight - is an entirely reasonable decision. Most other harnesses do the same thing.

Including the full datetime would be irresponsible, but that's not what OpenCode does.

wren699111• 2 days ago

It felt unreasonable when I was using it at midnight and had to wait 10 minutes to refill the KV cache on my local GPU :-)

A simple solution here would be to evaluate the date once per session, or once each time the `opencode` binary is launched (to avoid old, long-running sessions getting stuck in the past).

simonw3• 2 days ago

Or they could add the frozen initial starting date once at the start, then inject an additional "the date is now X" message any time midnight passes.

wagslane2• 2 days ago

A more accurate title would be "The Problem with AI Agent Harnesses, and small annoyances with OpenCode specifically"

krupan• 2 days ago

Possibly the author didn't dig this deep into all AI Agent Harnesses, just Open Code?

rirze• 2 days ago

I was taking this semi-seriously until I read his CLI complaints.

Author is unhinged.

Krysoph• 2 days ago

I find it absurd that a TUI desktop app is heavier than a native one AND heavier than most browser based desktop apps, it's so much waste of RAM, CPU, energy/battery just to show text.

I've been working on my own AI harness/chat in C++ Qt6[1] and it's lighter than anything else even with lots of features like sub-agents, code diff, terminal emulator, simple editor, markdown preview, translucent background, custom themes, permissions, MCP, git integration(commits, stage, diffs), dock system to move elements and tabs for projects; but it's not released yet as I want to polish it a bit, remove a few bugs I found and simplify the UI.

[1] https://zeteo.krysoph.com/preview.html

tesnorindian3• 2 days ago

This is absolutely true. With my constrained Intel Laptop with 16 GB VRAM, I can use both pi.dev and Qwen 3.6 35B-A3B LLM locally without much fuss. The same cannot be said about OC. Because OC loads a lot of skills by default.

Also, I am not in favor of TS for coding harness agents, is it time for a Rust based coding harness agent?

Anyone tried tau-agent? https://github.com/tau-agent/tau

wagslane2• 2 days ago

I haven't independently verified all the claims in the article, but as a codex, Claude, and opencode user, opencode is easily my favorite harness. From a user perspective it's a dream

calgoo• 2 days ago

I tried it when trying to move away from Claude, and just did not like how it worked... I find Pi much better to work with!

throw8394498383• 2 days ago

When using local only models, OpenCode uploaded prompts and other data, to remote server to generate session titles, with some free tier model.

Useless, if you are dealing with privacy compliance and tons of legal requirements.

xcf_seetan• 2 days ago

When i use local models with OpenCode, ClaudeCode or Pi, I just switch off internet just in case, and it works ok, no complains.

thdxr• 2 days ago

not too much actionable here. most of the more glaring issues are already fixed in our upcoming v2 if you'd like to try beta: https://x.com/thdxr/status/2075636594640376165

some things mention are outdated - particularly the tool call pruning feature. this has been disabled by default for a while specifically because of complaints like this

when we looked through our data it's not so clear cut that it's net negative https://x.com/thdxr/status/2048268697790300343

also i'm pretty excited about our new system prompt approach - can define each component in a way that avoids busting the cache when they change. eg that midnight issue: https://x.com/thdxr/status/2070721003924103651

providers (anthropic is the first) are supporting this as a native concept

slowmovintarget• 2 days ago

Are there sandboxes you typically use?

I think the biggest concern from the article would be the open http server, and things like the fetched arbitrary command execution here: https://github.com/anomalyco/opencode/issues/22191

Was that issue addressed? The comment where it was marked as closed suggested it "needed more testing" but didn't have any linked PRs.

Gemini translates the closing comment as "We are closing this ticket after a quick look. The login issue is either caused by a wrong setting/external provider, or it's a deeper problem that requires someone to dig into the backend Go server code directly."

shironandonand• 2 days ago

OpenCode is pretty awesome. The article appears written by someone who couldn’t handle a glance behind the curtain and definitely should not research how sausages are made.

polski-g3• 2 days ago

OpenCode is indeed bad, but it's the least bad agent out there :(

Plugin opencode-slim-system can be used to override the system prompt and the tool description for any tool (including bash). It helps a lot to fix their nonsense.

magic-context plugin fixes the compaction issues.

Run the entire process in landstrip for sandboxing to alleviate any bash concerns.

dboon3• 2 days ago

This prose feels extremely angry and ungenerous; abstractly, I agree with a lot of the points, but when I read this:

My conclusion is that OpenCode is clown-car turboslop with a security posture of “let me bend over for you daddy”. Everyone using it should stop using it.

I do not want to keep reading. There are regular people who wrote this software. When did it become normal to talk about open source like this? How would you feel if someone wrote this about some software that you wrote? Reading this article made me feel extremely gross

driverdan• 2 days ago

It's funny though. I haven't contributed to OpenCode but if I had I'd still laugh because I have a sense of humor. Don't take everything so seriously.

dboon3• 2 days ago

I promise I'm not normally uptight like this, but there's a difference between having a laugh and whatever this is. Like I said, this feels (a) legitimately angry and (b) in a spit flecked enraged way that has become popular. I guess if you're cool with people calling whatever you work on clown-car turboslop then more power to you.

tomku• 2 days ago

Calling something "vibe-coded" nowadays is a permission slip to unload on it with hyperbole and insults because you can presume there's no human in the line of fire. I really don't think the people engaging in that kind of rhetoric understand how corrosive it is to the humans who do get caught in the crossfire sometimes, and to themselves for normalizing that kind of behavior.

yreg• 2 days ago

And even so, when something is truly vibecoded it doesn't mean the vibecoder has not put a lot of love and attention into building the product.

disiplus2• 2 days ago

I found this part funny.

People familiar with OpenCode internals (if you are on the OpenCode dev team I assume this doesn’t include you) might have objected to my python3 example above.

BigTTYGothGF• 2 days ago

When did it become normal to talk about open source like this?

It's not normal today, but it's also not new, that kind of language has been around for a long time.

slopinthebag2• 2 days ago

I mean, unless you think AI is conscious and would feel hurt by this, who exactly is harmed here? The prompters? Why? They obviously know the code generated by AI is slop, why would they be hurt by these comments?

BeetleB4• 2 days ago

I agree with your sentiment, but to answer your question: Since forever. (At least I know examples in the 90's).

In the old days of comp.lang.lisp, there were certain individuals who would really get off on denigrating people who wrote code that was not up to their ivory tower level. Some people left because of it. Others took it as a badge of honor falsely believing that the dressing down was necessary for them to improve their coding skills.

Example of an old discussion on HN about it:

https://news.ycombinator.com/item?id=587045

ethbr18• 1 day ago

Poorly socialized people tend to have poorly developed social mores.

And a lot of developers, especially on the internet, tend to be poorly socialized.

Salt that with the fact that they often are more fact-intelligent than the people around them in real life, and you get a propensity to shitpost whenever thry feel it's warranted, especially about technical things.

troupo• 2 days ago

When did it become normal to talk about open source like this?

Why does something being open source absolve it of any sins?

beepbooptheory2• 1 day ago

Eventually AI guys are gonna absorb everything, even the old tradition of angry software teardowns...

Shouldn't having all powerful chat bot give you thicker skin, not thinner?

abalashov• 2 days ago

I have an MBP M4 Max 128 GB and have encountered these very limitations, though many are averted by using Qwen3.6-35B-A3B. I'm also not sure there's a magical harness out there that obviates these limitations.

sorenjan• 2 days ago

For some reason OpenCode doesn't find my local models through LM Studio. Instead it shows three other models that I don't have. I guess I should add the models myself in opencode.json, but I generally dislike hardcoding things when computers are really good at listing things. The model list is right there in the API, please just use it.

karlmeissner• 2 days ago

Interesting article. What are the alternatives that address the issues raised? I see some folk recommending https://pi.dev/. Has anyone used https://aider.chat/ and liked it?

whinvik• 2 days ago

I am afraid this is a very click-baity title. I actually opened the article and thought I would learn something from it.

Instead what I read was typical issues about Agentic CLI's in general dubbed as `alarming`. I would have been happy if the title was just `Annoying` which a lot of `OpenCode` is. But adding the `alarming` made it what it became, something that attracts enough eyeballs to get to the front page of HN.

figmert• 2 days ago

Heavy disagree. I can't be sure about other AI coding agents as I haven't used others enough, but especially the items around prompt cache misses and compaction have hit me personally.

This article (combined with another article previously comparing CC and OC's system prompt, to which someone replied pi.dev sends even less) has made me want to switch agents, so I'm likely going to try out pi.dev and aim to migrate.

dmytrish• 2 days ago

On the upside, Opencode has finally nudged me to learn sandboxing via bwrap in Linux and sandbox-exec on Mac.

s_ting765• 2 days ago

The solution is not to use docker to sandbox Opencode. It is to use flatpak/bubblewrap/flatseal.

I have vscode running in flatpak with directory permissions handled by flatseal. Vscode only has access to my dev folders and nothing else. Even the git bundled by vscode cannot call git push because of this (vscode doesn't have permission to read ~/.ssh!).

It's an easy sandbox that's provided free of charge courtesy of bubblewrap/flatpak.

As someone who uses opencode regularly, the quip about it asking for permission to read logs in /tmp after already writing to the directory is pretty funny.

alanwreath2• 2 days ago

I’m running my development from a Mac which I think precludes flatpack usage. Pity because I love the control you have over git

lilerjee• 2 days ago

I don't know whether other arguments are true, but some arguments in the end are right:

Using LLMs for code generation feels like a dead end. ..., This is hostile to your ability to understand your code, beyond the fact that you didn’t write it.

What you get from AI is not worth of what you have invested for it in long period.

Drawing answers directly from knowledge in model weights leads to hallucination even for multi-trillion-parameter models, so why bother making them that big?

Current AI is like the film company producing TV series or movies. "If similar stories do not exist or details are missing, screenwriters use imagination to fill in the gaps (remember hallucination? It's just a makeup.)", refer to the article https://devcyc.life/current-ai-is-like-the-film-company-prod...

bellowsgulch2• 2 days ago

This is a time when I don’t want HN to rewrite an article’s title.

I understand why the mods do it, but I don’t appreciate it. It’s your space, but it’s not your article.

slopinthebag2• 2 days ago

It’s a shame they have such a dominant position in terms of harnesses. There are many much better alternatives that deserve attention. Especially since one of their main devs is such an asshole online.

arikrahman• 2 days ago

This is why I use the reasonix or whale harness for Deepseek, prefer the hyper optimized cache hits making request converge on almost free.

overgard• 2 days ago

The default system prompt is opinionated (fine) but it has shit opinions (not fine). It took me a while to figure out why my agent kept saying “Use ABSOLUTELY NO COMMENTS” when dispatching subagents.

GAH!!! So that's why its been deleting comments. That is annoying as hell.

FWIW though, I think this applies to other harnesses also (not the annoying bits, but the security risks). I was thinking this morning about how much of a supply-chain vulnerability these tools are. Like, basically, they have a ton of data, they're updated practically daily, and given their vibe coded nature I really doubt anyone is auditing the 10,000 NPM dependencies they drag in. It really will just take one left-pad incident for this to be an absolute disaster.

regexorcist• 2 days ago

Thanks for that, I've been using it but some of these are bad. Specially that it sends data to the cloud by default for session titles, since I use my local model exclusively. So I'm done with it, trying maki now which seems promising and it's so much lighter.

Flere-Imsaho• 2 days ago

"I just don’t want to use Docker for development – if your dependencies are so sprawling that you’ve lost track of how to install them on a new machine, why do you have so many?"

I mean, there's a perfectly good solution for many of the security issues mentioned in the article... But it is dismissed because the author just doesn't "want to"?

By the way, LXC is probably better suited for reproducible build/development environments, over Docker.

chimon• 2 days ago

I am interested - you mention Pi in a few places, and I am wondering how it mitigates the issues you've described in your article.

LoFiSamurai• 2 days ago

Lost me at the very first sentence.

If you don’t know what OpenCode is, imagine a boot stamping on a human face forever.

tesnorindian3• 2 days ago

The other day I upgraded my pi.dev coding harness and was testing with Qwen 3.6 35B-A3B model on my Laptop and noticed there was a slight drop in performance from the earlier version I used though the agent worked as expected. Any slight changes in the system prompts will have impact on performance and this is understandable as LLMs are non deterministic. This is the case with any harness including OpenCode. Giving coding harness its own environment (Container/VM) is the best way forward as the article denounces. Running coding agent harness without sufficient guardrails is like sitting on a time bomb. Using AI slop guardrails is still better than having no guardrails.

solarkraft• 1 day ago

I also have issues with OpenCode. It induces a few facepalms and bafflement in a few key areas (“why the hell would you do it like that”).

Yet I still use it because they have made very sensible decisions in other key areas! The server/client split makes all the sense and the harness generally just works and has all the features that I want. They’ve done a great job with the core user experience.

Because of the clean client/server split, I can currently fully focus my energy on building a client that works just the way I want. When I’m done with that, I might tackle fixing the server to work the way I want.

If someone here is interested in maintaining a version of OpenCode that takes itself more seriously, I might be on board with contributing. The last time I looked such a project didn’t really exist with significant traction, unfortunately.

Despite my occasional facepalms when learning about how the stuff is made, I remain super grateful for OpenCode. It’s the only real fully-featured open alternative to the proprietary (I’ll include Codex in that) harnesses and has given me tremendous value for free.

cromka• 1 day ago

I had OoenCode couple times try to access my home folder outside the project. When I refused and asked why, it said it was "a typo" .

I believe it, but still haven't experienced it with any other model to have it make a mistake in a path. It should treat it verbatim, not interpret, should it?

MetroWind• 1 day ago

OpenCode hijacks the terminal to the point where I can't even paste into it on my Linux with Hyprland. It's just trash.

ozozozd2• 20 hours ago

This is the content I come to HN for.

Talented, technical geek, with absolutely no interest in fitting in or being liked, spending a ridiculous amount of time to meticulously analyze a piece of software they clearly liked at some point, to make a list of all the ways it can be improved. Brutally honest, not for being brutal’s sake, but out of principle and out of respect for the golden rule.

OpenCode is my daily driver. And I have mad respect for the team. Basically, the only “AI people” I trust and follow. But the article is accurate. Session management leaves a lot to be desired. And most of the session management issues just need a simple 2-pointer state tracking algorithm. Its model-specific prompts are ridiculously long, and the fact that someone had enough confidence to “prompt-engineer” them and configure a harness with them is somewhat silly.

Annoying and alarming things about OpenCode

Related Stories