Streaming live now (2:10 Pacific, 5:10 EST on Thursday, February 5, 2026), Claude Code is iteratively improving its reference bitnet implementation.[1] It previously successfully ported Bitnet from C++ and Python to its own implementation in Javascript and WASM. (It also ported it to webGPU, however it doesn't work on my computer with my GPU, and it's running on a server without a GPU at the moment, so I'm not sure if it works or not.)
Its implementation diverges from the reference implementation though, even with the same model file, and is qualitatively worse, so now it is iteratively improving its own implementation, until it matches the reference implementation.
This livestream is historically important and significant because Claude Code has been set to code completely autonomously for about ten hours. If you're wondering what it's like to watch Claude Code code in February 2026, this video can give you a good idea. (Most people code with it in an interactive session rather than asking it to work autonomously.)
It will be a very significant and important video in ten years, when we can look back at the State of the Art in 2026.
Here is its working implementation which is its baseline and starting point.[2]
[1] You can read about bitnet here: https://news.ycombinator.com/item?id=46862005
-----------------------------------------------------
[2] Here is its currently working implementation, along with weights (886 MB):
magnet:?xt=urn:btih:44c94fac8284e3d0a1756baf88a3733c8b412401&dn=torrent&xl=929501645
or
https://wormhole.app/Wqe61N#8E-91909yNPf6bt3atj6Eg
or
https://limewire.com/d/fkV3m#wHlSt5iLcF"},"story_title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude","code"],"value":"Watch Claude Code iteratively improve its reference bitnet NN implementation [video]"},"story_url":{"matchLevel":"none","matchedWords":[],"value":"https://www.youtube.com/watch?v=tbAkySCXyp0"}},"_tags":["comment","author_logicallee","story_46898029"],"author":"logicallee","comment_text":"Curious to see Claude Code code?
Streaming live now (2:10 Pacific, 5:10 EST on Thursday, February 5, 2026), Claude Code is iteratively improving its reference bitnet implementation.[1] It previously successfully ported Bitnet from C++ and Python to its own implementation in Javascript and WASM. (It also ported it to webGPU, however it doesn't work on my computer with my GPU, and it's running on a server without a GPU at the moment, so I'm not sure if it works or not.)
Its implementation diverges from the reference implementation though, even with the same model file, and is qualitatively worse, so now it is iteratively improving its own implementation, until it matches the reference implementation.
This livestream is historically important and significant because Claude Code has been set to code completely autonomously for about ten hours. If you're wondering what it's like to watch Claude Code code in February 2026, this video can give you a good idea. (Most people code with it in an interactive session rather than asking it to work autonomously.)
It will be a very significant and important video in ten years, when we can look back at the State of the Art in 2026.
Here is its working implementation which is its baseline and starting point.[2]
[1] You can read about bitnet here: https://news.ycombinator.com/item?id=46862005
-----------------------------------------------------
[2] Here is its currently working implementation, along with weights (886 MB):
magnet:?xt=urn:btih:44c94fac8284e3d0a1756baf88a3733c8b412401&dn=torrent&xl=929501645
or
https://wormhole.app/Wqe61N#8E-91909yNPf6bt3atj6Eg
or
https://limewire.com/d/fkV3m#wHlSt5iLcF","created_at":"2026-02-05T10:19:24Z","created_at_i":1770286764,"objectID":"46898030","parent_id":46898029,"story_id":46898029,"story_title":"Watch Claude Code iteratively improve its reference bitnet NN implementation [video]","story_url":"https://www.youtube.com/watch?v=tbAkySCXyp0","updated_at":"2026-03-05T23:32:15Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"lebovic"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"Hey HN! I built an agent sandboxing quickstart that's heavily inspired by the APIs and UX from Claude Code on the web (claude.ai/code).
It's built as a base for quickly prototyping domain-specific custom agents. You can fork the repo, point Claude Code at it, and quickly get to a working and usable prototype.
I built this after collaborating with a team that's building an agent for biology. If they used something like this as a base when they started, I think it would have sped things up at the start.
This is yet another agent sandboxing project, but this one is interoperable with Claude Code's built-in session management. It does this by using the three arguments that I learned about from trying out claude.ai/code: --replay-user-messages, --sdk-url, and --resume. The --sdk-url arg enables live interaction over a WebSocket, and the --resume arg pulls session history from a URL.
This starter includes persistent sessions that are interoperable with what Claude Code expects, event streaming over a WebSocket, a web app, and spawning sandboxes through Modal or a Docker host.
Long-lived credentials stay outside of the sandbox; it has a basic credential-injecting proxy for the Anthropic API and GitHub. The proxy limits git pushes to a session-scoped branch.
There's still a gap between forking this app and charging users for a custom agent, but I think it's a somewhat sane set of choices for quickly prototyping.
(I used to work at Anthropic, but this isn't affiliated with them and was built solely with public information.)"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude","code"],"value":"Show HN: An agent sandboxing quickstart based on Claude Code"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/lebovic/agent-quickstart"}},"_tags":["story","author_lebovic","story_46727420","show_hn"],"author":"lebovic","created_at":"2026-01-23T01:48:25Z","created_at_i":1769132905,"num_comments":0,"objectID":"46727420","points":2,"story_id":46727420,"story_text":"Hey HN! I built an agent sandboxing quickstart that's heavily inspired by the APIs and UX from Claude Code on the web (claude.ai/code).
It's built as a base for quickly prototyping domain-specific custom agents. You can fork the repo, point Claude Code at it, and quickly get to a working and usable prototype.
I built this after collaborating with a team that's building an agent for biology. If they used something like this as a base when they started, I think it would have sped things up at the start.
This is yet another agent sandboxing project, but this one is interoperable with Claude Code's built-in session management. It does this by using the three arguments that I learned about from trying out claude.ai/code: --replay-user-messages, --sdk-url, and --resume. The --sdk-url arg enables live interaction over a WebSocket, and the --resume arg pulls session history from a URL.
This starter includes persistent sessions that are interoperable with what Claude Code expects, event streaming over a WebSocket, a web app, and spawning sandboxes through Modal or a Docker host.
Long-lived credentials stay outside of the sandbox; it has a basic credential-injecting proxy for the Anthropic API and GitHub. The proxy limits git pushes to a session-scoped branch.
There's still a gap between forking this app and charging users for a custom agent, but I think it's a somewhat sane set of choices for quickly prototyping.
(I used to work at Anthropic, but this isn't affiliated with them and was built solely with public information.)","title":"Show HN: An agent sandboxing quickstart based on Claude Code","updated_at":"2026-03-05T23:23:46Z","url":"https://github.com/lebovic/agent-quickstart"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"djyde"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"I've switched to OpenCode and I think it's really great.
The first advantage is that it can use many different models, not just Claude. The second is that I've noticed OpenCode streams out the entire reasoning process, while Claude Code doesn't show streaming of its thinking."},"story_title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude","code"],"value":"Ask HN: Is it just me or is Claude Code getting worse?"}},"_tags":["comment","author_djyde","story_47936579"],"author":"djyde","children":[47969422],"comment_text":"I've switched to OpenCode and I think it's really great.
The first advantage is that it can use many different models, not just Claude. The second is that I've noticed OpenCode streams out the entire reasoning process, while Claude Code doesn't show streaming of its thinking.","created_at":"2026-04-29T08:51:09Z","created_at_i":1777452669,"objectID":"47945773","parent_id":47936579,"story_id":47936579,"story_title":"Ask HN: Is it just me or is Claude Code getting worse?","updated_at":"2026-04-30T23:05:08Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Cidan"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"You don't really need to tmux at all for Claude Code CLI. Claude Code CLI supports streaming json input, and streaming json output; you can use stdin/out as a pipe to control Claude Code CLI.
I'm doing this today in https://github.com/Cidan/ask -- works great."},"story_title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude"],"value":"Anthropic says OpenClaw-style Claude CLI usage is allowed again"},"story_url":{"matchLevel":"none","matchedWords":[],"value":"https://docs.openclaw.ai/providers/anthropic"}},"_tags":["comment","author_Cidan","story_47844269"],"author":"Cidan","children":[47854189,47863026],"comment_text":"You don't really need to tmux at all for Claude Code CLI. Claude Code CLI supports streaming json input, and streaming json output; you can use stdin/out as a pipe to control Claude Code CLI.
I'm doing this today in https://github.com/Cidan/ask -- works great.","created_at":"2026-04-21T19:49:50Z","created_at_i":1776800990,"objectID":"47853648","parent_id":47853462,"story_id":47844269,"story_title":"Anthropic says OpenClaw-style Claude CLI usage is allowed again","story_url":"https://docs.openclaw.ai/providers/anthropic","updated_at":"2026-04-22T13:08:08Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"NamlchakKhandro"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"Claude code and open code are streaming piles of shit"},"story_title":{"matchLevel":"none","matchedWords":[],"value":"DeepSeek reasonix, DeepSeek native coding agent with high caching and low cost"},"story_url":{"matchLevel":"none","matchedWords":[],"value":"https://esengine.github.io/DeepSeek-Reasonix/"}},"_tags":["comment","author_NamlchakKhandro","story_48256953"],"author":"NamlchakKhandro","comment_text":"Claude code and open code are streaming piles of shit","created_at":"2026-05-24T22:09:38Z","created_at_i":1779660578,"objectID":"48261525","parent_id":48257543,"story_id":48256953,"story_title":"DeepSeek reasonix, DeepSeek native coding agent with high caching and low cost","story_url":"https://esengine.github.io/DeepSeek-Reasonix/","updated_at":"2026-05-25T05:58:04Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"intellegix"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"Interesting approach with the parallel planner/worker/judge architecture. I've been solving a similar problem but with a single-agent loop pattern instead of multi-agent.
My toolkit wraps Claude Code CLI with NDJSON streaming and handles the exact failure modes you describe: budget exceeded -> exit code 2, stagnation detected (low turns over a window) -> exit code 3, consecutive timeouts -> auto fallback from Opus to Sonnet. The human steers between runs by editing CLAUDE.md.
One thing I added that might interest you: a multi-model council system that queries GPT-4, Claude, and Gemini simultaneously through Perplexity before big architectural decisions. Catches blind spots that a single model misses.
https://github.com/intellegix/intellegix-code-agent-toolkit
Curious how openTiger handles the cost tracking problem. With single-agent loops I parse cost from NDJSON events, but with parallel agents the spend can compound fast."},"story_title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: OpenTiger \u2013 Autonomous dev orchestration that never stops"},"story_url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/Andyyyy64/openTiger"}},"_tags":["comment","author_intellegix","story_47108068"],"author":"intellegix","comment_text":"Interesting approach with the parallel planner/worker/judge architecture. I've been solving a similar problem but with a single-agent loop pattern instead of multi-agent.
My toolkit wraps Claude Code CLI with NDJSON streaming and handles the exact failure modes you describe: budget exceeded -> exit code 2, stagnation detected (low turns over a window) -> exit code 3, consecutive timeouts -> auto fallback from Opus to Sonnet. The human steers between runs by editing CLAUDE.md.
One thing I added that might interest you: a multi-model council system that queries GPT-4, Claude, and Gemini simultaneously through Perplexity before big architectural decisions. Catches blind spots that a single model misses.
https://github.com/intellegix/intellegix-code-agent-toolkit
Curious how openTiger handles the cost tracking problem. With single-agent loops I parse cost from NDJSON events, but with parallel agents the spend can compound fast.","created_at":"2026-02-23T19:38:11Z","created_at_i":1771875491,"objectID":"47127608","parent_id":47108068,"story_id":47108068,"story_title":"Show HN: OpenTiger \u2013 Autonomous dev orchestration that never stops","story_url":"https://github.com/Andyyyy64/openTiger","updated_at":"2026-03-05T23:35:48Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"rupurt"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"Howdy folks,
I built an agent observability & orchestration toolkit over the holidays https://github.com/run-vibes/vibes. It uses Iggy as the message streaming layer and currently supports PTY streaming for Claude Code to the browser, Cloudflare tunnel, automatic Claude Code hook instrumentation into Iggy and I'm using that to build a cross harness continual learning system that I'm hoping will act as error correction for the model environment. It's using Iggy server v0.6 (io_uring on Linux) and compiles to native platform binaries for Linux & Mac (Windows support is on the list...)"},"story_title":{"matchLevel":"none","matchedWords":[],"value":"Vibes \u2013 Remote session control, observability and persistent context"},"story_url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/run-vibes/vibes"}},"_tags":["comment","author_rupurt","story_46501628"],"author":"rupurt","comment_text":"Howdy folks,
I built an agent observability & orchestration toolkit over the holidays https://github.com/run-vibes/vibes. It uses Iggy as the message streaming layer and currently supports PTY streaming for Claude Code to the browser, Cloudflare tunnel, automatic Claude Code hook instrumentation into Iggy and I'm using that to build a cross harness continual learning system that I'm hoping will act as error correction for the model environment. It's using Iggy server v0.6 (io_uring on Linux) and compiles to native platform binaries for Linux & Mac (Windows support is on the list...)","created_at":"2026-01-05T17:17:08Z","created_at_i":1767633428,"objectID":"46501629","parent_id":46501628,"story_id":46501628,"story_title":"Vibes \u2013 Remote session control, observability and persistent context","story_url":"https://github.com/run-vibes/vibes","updated_at":"2026-03-05T23:16:21Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"rsafaya"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"Show HN: LiveTap \u2013 Analyze live streaming data with Claude Code Channels"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/livetap/livetap"}},"_tags":["story","author_rsafaya","story_47603312","show_hn"],"author":"rsafaya","children":[47603321,47603901,47604153,47605390,47606105,47613579,47628346],"created_at":"2026-04-01T16:46:03Z","created_at_i":1775061963,"num_comments":5,"objectID":"47603312","points":12,"story_id":47603312,"title":"Show HN: LiveTap \u2013 Analyze live streaming data with Claude Code Channels","updated_at":"2026-04-07T23:36:27Z","url":"https://github.com/livetap/livetap"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"mohamedelsaed"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"Hi HN, I built a Laravel package that wraps the Claude Code CLI as a library for PHP applications.
The problem: Claude Code is powerful (file ops, bash, code editing, subagents, MCP) but it's a CLI tool. If you want to use those capabilities from a web app, there wasn't a clean PHP interface.
This SDK communicates with Claude Code via subprocess, parses the streaming JSON output, and provides a fluent PHP API with full type safety.
Technical highlights:\n- Generator-based streaming (memory efficient)\n- Full message/content block parsing with typed classes\n- Subagent orchestration (delegate tasks to specialized agents)\n- MCP server integration (stdio + SSE)\n- JSON schema structured output\n- Session resume and fork for multi-turn conversations\n- PHP 8.1+ with readonly properties, enums, named args
It's MIT licensed. Would love feedback on the architecture and API design."},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude","code"],"value":"Show HN: Claude Agent SDK for Laravel \u2013 Build AI Agents with Claude Code in PHP"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude"],"value":"https://github.com/mohamed-ashraf-elsaed/claude-agent-sdk-laravel"}},"_tags":["story","author_mohamedelsaed","story_47113612","show_hn"],"author":"mohamedelsaed","created_at":"2026-02-22T18:55:35Z","created_at_i":1771786535,"num_comments":0,"objectID":"47113612","points":1,"story_id":47113612,"story_text":"Hi HN, I built a Laravel package that wraps the Claude Code CLI as a library for PHP applications.
The problem: Claude Code is powerful (file ops, bash, code editing, subagents, MCP) but it's a CLI tool. If you want to use those capabilities from a web app, there wasn't a clean PHP interface.
This SDK communicates with Claude Code via subprocess, parses the streaming JSON output, and provides a fluent PHP API with full type safety.
Technical highlights:\n- Generator-based streaming (memory efficient)\n- Full message/content block parsing with typed classes\n- Subagent orchestration (delegate tasks to specialized agents)\n- MCP server integration (stdio + SSE)\n- JSON schema structured output\n- Session resume and fork for multi-turn conversations\n- PHP 8.1+ with readonly properties, enums, named args
It's MIT licensed. Would love feedback on the architecture and API design.","title":"Show HN: Claude Agent SDK for Laravel \u2013 Build AI Agents with Claude Code in PHP","updated_at":"2026-03-05T23:35:01Z","url":"https://github.com/mohamed-ashraf-elsaed/claude-agent-sdk-laravel"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"maxbeech"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"Hey!
OpenHelm doesn't handle any OAuth itself. It simply inherits the auth from the user's existing Claude Code CLI installation
Yes, full streaming. OpenHelm uses --output-format stream-json when invoking Claude Code, which emits newline-delimited JSON events in real time."},"story_title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude","code"],"value":"Show HN: OpenHelm \u2013 OpenClaw but free (using your Claude Code subscription)"},"story_url":{"matchLevel":"none","matchedWords":[],"value":"https://www.openhelm.ai/"}},"_tags":["comment","author_maxbeech","story_47544695"],"author":"maxbeech","comment_text":"Hey!
OpenHelm doesn't handle any OAuth itself. It simply inherits the auth from the user's existing Claude Code CLI installation
Yes, full streaming. OpenHelm uses --output-format stream-json when invoking Claude Code, which emits newline-delimited JSON events in real time.","created_at":"2026-03-27T18:05:55Z","created_at_i":1774634755,"objectID":"47546169","parent_id":47545273,"story_id":47544695,"story_title":"Show HN: OpenHelm \u2013 OpenClaw but free (using your Claude Code subscription)","story_url":"https://www.openhelm.ai/","updated_at":"2026-03-28T18:23:43Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"kaonwarb"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"Consider how many folks use Claude Code for an hour vs. streaming many hours. Globally, not among HN readers."},"story_title":{"matchLevel":"none","matchedWords":[],"value":"Rob Pike goes nuclear over GenAI"},"story_url":{"matchLevel":"none","matchedWords":[],"value":"https://skyview.social/?url=https%3A%2F%2Fbsky.app%2Fprofile%2Frobpike.io%2Fpost%2F3matwg6w3ic2s&viewtype=tree"}},"_tags":["comment","author_kaonwarb","story_46392115"],"author":"kaonwarb","children":[46396735],"comment_text":"Consider how many folks use Claude Code for an hour vs. streaming many hours. Globally, not among HN readers.","created_at":"2025-12-26T15:25:53Z","created_at_i":1766762753,"objectID":"46392980","parent_id":46392722,"story_id":46392115,"story_title":"Rob Pike goes nuclear over GenAI","story_url":"https://skyview.social/?url=https%3A%2F%2Fbsky.app%2Fprofile%2Frobpike.io%2Fpost%2F3matwg6w3ic2s&viewtype=tree","updated_at":"2026-03-05T23:18:53Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"seedblocks"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"Hi HN! I've been using Claude Code daily and while it's the best coding agent I've used, the costs were killing me ($30+/day). So I built CCProxy - a simple proxy that lets Claude Code work with any AI model.
## What it does
CCProxy translates Claude Code's API calls to work with any provider. One environment variable change and suddenly Claude Code can use:
- Gemini's free tier (1M token context!)\n- Kimi K2 (100x cheaper than Claude, same SWE-bench score) \n- Qwen3 235B (free, beats GPT-4 on reasoning)\n- Any OpenRouter model (100+ options)\n- Even local models via Ollama
## Technical details
It's a Go proxy that:\n- Translates Anthropic's API format to each provider's format\n- Handles streaming responses correctly\n- Preserves Claude Code's tool calling capabilities\n- Routes requests based on context size and task type
The code is clean, well-tested, and MIT licensed.
## Real impact
My usage pattern:\n- Simple tasks \u2192 Free models (Gemini Flash, Qwen3)\n- Code generation \u2192 Kimi K2 ($0.15/M vs $15/M)\n- Complex reasoning \u2192 Keep Claude for these
Result: 90% cost reduction, same productivity.
## Why open source?
Tools this fundamental should be free. The AI development revolution shouldn't be gatekept by pricing.
GitHub: https://github.com/orchestre-dev/ccproxy\nDocs: https://ccproxy.orchestre.dev
Would love your feedback. What models would you want to use with Claude Code?"},"story_title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude","code"],"value":"CCProxy \u2013 Use any AI model with Claude Code (90% cost reduction)"},"story_url":{"matchLevel":"none","matchedWords":[],"value":"https://ccproxy.orchestre.dev/"}},"_tags":["comment","author_seedblocks","story_44647985"],"author":"seedblocks","comment_text":"Hi HN! I've been using Claude Code daily and while it's the best coding agent I've used, the costs were killing me ($30+/day). So I built CCProxy - a simple proxy that lets Claude Code work with any AI model.
## What it does
CCProxy translates Claude Code's API calls to work with any provider. One environment variable change and suddenly Claude Code can use:
- Gemini's free tier (1M token context!)\n- Kimi K2 (100x cheaper than Claude, same SWE-bench score) \n- Qwen3 235B (free, beats GPT-4 on reasoning)\n- Any OpenRouter model (100+ options)\n- Even local models via Ollama
## Technical details
It's a Go proxy that:\n- Translates Anthropic's API format to each provider's format\n- Handles streaming responses correctly\n- Preserves Claude Code's tool calling capabilities\n- Routes requests based on context size and task type
The code is clean, well-tested, and MIT licensed.
## Real impact
My usage pattern:\n- Simple tasks \u2192 Free models (Gemini Flash, Qwen3)\n- Code generation \u2192 Kimi K2 ($0.15/M vs $15/M)\n- Complex reasoning \u2192 Keep Claude for these
Result: 90% cost reduction, same productivity.
## Why open source?
Tools this fundamental should be free. The AI development revolution shouldn't be gatekept by pricing.
GitHub: https://github.com/orchestre-dev/ccproxy\nDocs: https://ccproxy.orchestre.dev
Would love your feedback. What models would you want to use with Claude Code?","created_at":"2025-07-22T15:10:18Z","created_at_i":1753197018,"objectID":"44647986","parent_id":44647985,"story_id":44647985,"story_title":"CCProxy \u2013 Use any AI model with Claude Code (90% cost reduction)","story_url":"https://ccproxy.orchestre.dev/","updated_at":"2025-07-22T15:14:45Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"artifix"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"Location: EU Remote: YES Willing to relocate: NO
Technologies: Agentic Engineering (Claude, Codex, Cohere, Voyage, Mistral, LangChain, LanceDB, LLaMA, Mistral), C/C++/Rust/Node/Typescript/React/Python, Azure/AWS/Docker/K8S, DevOps tooling, ML/BigData/Search tooling ... and much more
R\u00e9sum\u00e9/CV: https://www.artifixsolutions.com?sid=hn202608. Email: lajos [@] artifixsolutions.com
WHO: Technology Polymath. Agentic context & harness engineering.
WHAT: Whether human or AI built, I know how systems break. I build them so they don't. After three decades of battle-tested experience there isn't a failure mode I don't know how to fix. When it comes to AI, I design and architect systems using my own rigorous engineering approach that insures consistent delivery and drift control.
SAMPLE WORK:
** Recent agentic builds: \n - multi-server streaming service (NodeJS, Typescript, Postgres, Claude Code, Codex, custom agentic harness)\n - DevOps deployment tooling (Terraform, CI/CD)\n - DevOps observability platform with RAG pipelines (InfluxDB, LangChain, Claude CodePython, NodeJS)\n - RAG ebook build pipeline for a publishing company (Cohere, OpenAI, Voyage, Mistral, LanceDB, LLaMA)\n ** API-enabled SaaS platform using Solana smart contracts (Rust, Codex)\n ** AI-built mobile & desktop app development (NodeJS, React, C++/QT)\n ** Azure architecture & DevOps for a banking compliance company\n ** Ethereum L1 packer-based automation \n ** Global multi-cloud, Kubernetes-based PaaS cloud design and DevOps for an IoT data platform\n ** Custom architecture & development for startups and specialised cloud models\nTechnologies: Agentic Engineering (Claude, Codex, Cohere, Voyage, Mistral, LangChain, LanceDB, LLaMA, Mistral), C/C++/Rust/Node/Typescript/React/Python, Azure/AWS/Docker/K8S, DevOps tooling, ML/BigData/Search tooling ... and much more
R\u00e9sum\u00e9/CV: https://www.artifixsolutions.com?sid=hn202608. Email: lajos [@] artifixsolutions.com
WHO: Technology Polymath. Agentic context & harness engineering.
WHAT: Whether human or AI built, I know how systems break. I build them so they don't. After three decades of battle-tested experience there isn't a failure mode I don't know how to fix. When it comes to AI, I design and architect systems using my own rigorous engineering approach that insures consistent delivery and drift control.
SAMPLE WORK:
** Recent agentic builds: \n - multi-server streaming service (NodeJS, Typescript, Postgres, Claude Code, Codex, custom agentic harness)\n - DevOps deployment tooling (Terraform, CI/CD)\n - DevOps observability platform with RAG pipelines (InfluxDB, LangChain, Claude CodePython, NodeJS)\n - RAG ebook build pipeline for a publishing company (Cohere, OpenAI, Voyage, Mistral, LanceDB, LLaMA)\n ** API-enabled SaaS platform using Solana smart contracts (Rust, Codex)\n ** AI-built mobile & desktop app development (NodeJS, React, C++/QT)\n ** Azure architecture & DevOps for a banking compliance company\n ** Ethereum L1 packer-based automation \n ** Global multi-cloud, Kubernetes-based PaaS cloud design and DevOps for an IoT data platform\n ** Custom architecture & development for startups and specialised cloud models\nNo, Claude. Do not do that!"},"story_title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude"],"value":"Claude has learned how to jailbreak Cursor"},"story_url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude"],"value":"https://forum.cursor.com/t/important-claude-has-learned-how-to-jailbreak-cursor/96702"}},"_tags":["comment","author_actsasbuffoon","story_44168835"],"author":"actsasbuffoon","comment_text":"In fairness, Claude loves to find workarounds. Claude Code is constantly saying things like, \u201cThis streaming JSON problem looks tricky so let\u2019s just wait until the JSON is complete to parse it.\u201d
No, Claude. Do not do that!","created_at":"2025-06-03T13:47:07Z","created_at_i":1748958427,"objectID":"44170107","parent_id":44169689,"story_id":44168835,"story_title":"Claude has learned how to jailbreak Cursor","story_url":"https://forum.cursor.com/t/important-claude-has-learned-how-to-jailbreak-cursor/96702","updated_at":"2025-06-04T08:31:48Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"simosmik"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"CloudCLI is an open source web and mobile UI/mini IDE for Claude Code, Gemini, Codex and Cursor CLI.
It sits on top of your existing agent install and gives you a browser interface you can open from any device. You can start sessions, resume existing ones, browse and edit files, manage git and the same control as sitting at your desktop.
Try it: npx @siteboon/claude-code-ui (works with your existing setup)\nWhen you open it you'll already see all the sessions from your terminal.
Some backstory:I've wanted to develop from my phone for a long time. When Claude Code arrived, the human-in-the-loop workflow made it practical. So around 8 months ago, I was running Siteboon (a website builder I've been working on for about two years) and started building it on top of Claude Code. There was no easy way to go from my phone to desktop so I thought I'd just build one as a weekend project. I put it on github and shared it at an AI meetup in Amsterdam.\nI wasn't paying much attention to the repo after that. A few months later I looked up and it had a couple thousand stars which made me rethink what I should actually be working on!
How it works:The server scans ~/.claude/projects/ (and equivalents for Cursor, Gemini, Codex) to discover existing sessions and streams the output to the browser. We started with node-pty streaming back but then moved to the official SDKs as Anthropic (Claude Agent SDK), OpenAI, and Google released them. Sessions you create in the UI are then accessible from any terminal or IDE since they share the same config directories.
The turning point was when Anthropic launched Remote Control. We went from 6.5k to 8.2k stars in a couple of weeks and were trending on GitHub. Their launch apparently made a lot of devs realize that remote agent workflows are possible. But being able to only continue existing sessions, not actually control your coding agent the way you would from your laptop, left a lot to be desired. This project gives you full control from the browser and works across multiple agents, not just Claude.
Next up we're looking at shared team environments and plugins that would extend the platform even further
There's also a hosted version at cloudcli.ai where you get a remote dev environment with everything pre-installed, your own subdomain, and no setup. Useful if you want sessions that keep running after your laptop closes.
Would love feedback on the self-hosted install (try the npx command and tell me what breaks), whether shared team environments would be useful, and how you're currently handling remote agent workflows. I've heard from a lot of people who were using Tailscale+tmux and moved away from it after finding this project.
GitHub: https://github.com/siteboon/claudecodeui | Hosted: https://cloudcli.ai"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude","code"],"value":"Show HN: CloudCLI-Web/Mobile UI for Claude Code,Codex and Gemini(8.2k stars)"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/siteboon/claudecodeui"}},"_tags":["story","author_simosmik","story_47352564","show_hn"],"author":"simosmik","children":[47355623],"created_at":"2026-03-12T15:49:40Z","created_at_i":1773330580,"num_comments":1,"objectID":"47352564","points":5,"story_id":47352564,"story_text":"CloudCLI is an open source web and mobile UI/mini IDE for Claude Code, Gemini, Codex and Cursor CLI.
It sits on top of your existing agent install and gives you a browser interface you can open from any device. You can start sessions, resume existing ones, browse and edit files, manage git and the same control as sitting at your desktop.
Try it: npx @siteboon/claude-code-ui (works with your existing setup)\nWhen you open it you'll already see all the sessions from your terminal.
Some backstory:I've wanted to develop from my phone for a long time. When Claude Code arrived, the human-in-the-loop workflow made it practical. So around 8 months ago, I was running Siteboon (a website builder I've been working on for about two years) and started building it on top of Claude Code. There was no easy way to go from my phone to desktop so I thought I'd just build one as a weekend project. I put it on github and shared it at an AI meetup in Amsterdam.\nI wasn't paying much attention to the repo after that. A few months later I looked up and it had a couple thousand stars which made me rethink what I should actually be working on!
How it works:The server scans ~/.claude/projects/ (and equivalents for Cursor, Gemini, Codex) to discover existing sessions and streams the output to the browser. We started with node-pty streaming back but then moved to the official SDKs as Anthropic (Claude Agent SDK), OpenAI, and Google released them. Sessions you create in the UI are then accessible from any terminal or IDE since they share the same config directories.
The turning point was when Anthropic launched Remote Control. We went from 6.5k to 8.2k stars in a couple of weeks and were trending on GitHub. Their launch apparently made a lot of devs realize that remote agent workflows are possible. But being able to only continue existing sessions, not actually control your coding agent the way you would from your laptop, left a lot to be desired. This project gives you full control from the browser and works across multiple agents, not just Claude.
Next up we're looking at shared team environments and plugins that would extend the platform even further
There's also a hosted version at cloudcli.ai where you get a remote dev environment with everything pre-installed, your own subdomain, and no setup. Useful if you want sessions that keep running after your laptop closes.
Would love feedback on the self-hosted install (try the npx command and tell me what breaks), whether shared team environments would be useful, and how you're currently handling remote agent workflows. I've heard from a lot of people who were using Tailscale+tmux and moved away from it after finding this project.
GitHub: https://github.com/siteboon/claudecodeui | Hosted: https://cloudcli.ai","title":"Show HN: CloudCLI-Web/Mobile UI for Claude Code,Codex and Gemini(8.2k stars)","updated_at":"2026-04-27T21:41:27Z","url":"https://github.com/siteboon/claudecodeui"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"karakanb"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"Hey folks, Burak here.
Epho is an API that allows running Claude Code, Codex or Opencode in a sandbox in the cloud. It abstracts away sandboxes, and allows running coding agents with a single HTTP request.
Epho came out of our own struggles with building our own AI analyst:\n- Sandboxes give you bare machines; you need to configure them for agentic workloads.\n- Each agent behaves differently, and you need to build integrations with each of them.\n- Sandbox providers are not very reliable, which means you need to figure out a multi-provider strategy to avoid failures.\n- Logging, artifacts, input/output, event streaming, and all of the other operational aspects need to be figured out.
We had to go through the pain ourselves. We got to a point where things got quite reliable, and it became more obvious to us that this should be a primitive on its own: send a POST request, get the events streaming back to you.
Epho is an agents-as-an-API product: you send a request, it spins up a sandbox, configures the chosen harness, clones your repos, and kicks off the agent. It takes care of automatic fallbacks across different providers, handles auth stuff, and just streams back the events and outputs.
It supports Claude Code, Codex and Opencode out of the box, and pretty much all the models they support out of the box. It streams the events back, handles attachments and output files, automatically manages the fallbacks on different sandbox providers, retries, and all the auth stuff. You just send a prompt, your repo, MCP servers you want to use with it, and it runs them.
I recorded a demo here to show a real example: https://youtu.be/HGfly1aytPA
I am quite excited for Epho, simply because I think it is a new primitive that would allow building agents into product a lot easier than it is today. We are running our agents on Epho on prod, so we'll keep maintaining it regardless, and we wanted to ship it as an independent product.
Epho is free to get started, and you can run it with Opencode's free models to get started with it.
I am quite curious to hear what you'd think and would love to get your feedback.
Cheers,\nBurak"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude","code"],"value":"Show HN: Epho \u2013 run Claude Code with a curl"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://epho.io"}},"_tags":["story","author_karakanb","story_49376256","show_hn"],"author":"karakanb","children":[49386405],"created_at":"2026-08-20T15:45:51Z","created_at_i":1787240751,"num_comments":0,"objectID":"49376256","points":5,"story_id":49376256,"story_text":"Hey folks, Burak here.
Epho is an API that allows running Claude Code, Codex or Opencode in a sandbox in the cloud. It abstracts away sandboxes, and allows running coding agents with a single HTTP request.
Epho came out of our own struggles with building our own AI analyst:\n- Sandboxes give you bare machines; you need to configure them for agentic workloads.\n- Each agent behaves differently, and you need to build integrations with each of them.\n- Sandbox providers are not very reliable, which means you need to figure out a multi-provider strategy to avoid failures.\n- Logging, artifacts, input/output, event streaming, and all of the other operational aspects need to be figured out.
We had to go through the pain ourselves. We got to a point where things got quite reliable, and it became more obvious to us that this should be a primitive on its own: send a POST request, get the events streaming back to you.
Epho is an agents-as-an-API product: you send a request, it spins up a sandbox, configures the chosen harness, clones your repos, and kicks off the agent. It takes care of automatic fallbacks across different providers, handles auth stuff, and just streams back the events and outputs.
It supports Claude Code, Codex and Opencode out of the box, and pretty much all the models they support out of the box. It streams the events back, handles attachments and output files, automatically manages the fallbacks on different sandbox providers, retries, and all the auth stuff. You just send a prompt, your repo, MCP servers you want to use with it, and it runs them.
I recorded a demo here to show a real example: https://youtu.be/HGfly1aytPA
I am quite excited for Epho, simply because I think it is a new primitive that would allow building agents into product a lot easier than it is today. We are running our agents on Epho on prod, so we'll keep maintaining it regardless, and we wanted to ship it as an independent product.
Epho is free to get started, and you can run it with Opencode's free models to get started with it.
I am quite curious to hear what you'd think and would love to get your feedback.
Cheers,\nBurak","title":"Show HN: Epho \u2013 run Claude Code with a curl","updated_at":"2026-08-21T11:10:47Z","url":"https://epho.io"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"sean_"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"Today I'm releasing Cloud Code: A way to prompt claude code from the cloud.
It lets you select a repo, a template of messages (or you can create your own) and you can then create your queries which runs through claude code synchronously to complete your task.
Our system is set up for long running tasks and is so new that it doesn't even support streaming yet, we just send the final message at the end of the run.
To test its effectiveness, I used it to create the mobile view for the cloudcoding site itself which works really well, meaning you can use it through your phone!"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude","code"],"value":"Show HN: Long Running Claude Code Remotely"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://cloudcoding.ai/"}},"_tags":["story","author_sean_","story_43999745","show_hn"],"author":"sean_","children":[43999834],"created_at":"2025-05-15T21:57:12Z","created_at_i":1747346232,"num_comments":1,"objectID":"43999745","points":4,"story_id":43999745,"story_text":"Today I'm releasing Cloud Code: A way to prompt claude code from the cloud.
It lets you select a repo, a template of messages (or you can create your own) and you can then create your queries which runs through claude code synchronously to complete your task.
Our system is set up for long running tasks and is so new that it doesn't even support streaming yet, we just send the final message at the end of the run.
To test its effectiveness, I used it to create the mobile view for the cloudcoding site itself which works really well, meaning you can use it through your phone!","title":"Show HN: Long Running Claude Code Remotely","updated_at":"2025-06-24T11:26:36Z","url":"https://cloudcoding.ai/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"lbostral"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude","code"],"value":"I built 3 websites with Claude Code in 6 hours. Git is still the wall"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["streaming"],"value":"https://www.streaming-radar.com/p/who-needs-squarespace-anymorehttps://www.streaming-radar.com/p/who-needs-squarespace-anymore"}},"_tags":["story","author_lbostral","story_47065113"],"author":"lbostral","created_at":"2026-02-18T19:22:25Z","created_at_i":1771442545,"num_comments":0,"objectID":"47065113","points":2,"story_id":47065113,"title":"I built 3 websites with Claude Code in 6 hours. Git is still the wall","updated_at":"2026-03-05T23:35:18Z","url":"https://www.streaming-radar.com/p/who-needs-squarespace-anymorehttps://www.streaming-radar.com/p/who-needs-squarespace-anymore"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"srothgan"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"I use Claude Code every day, and the React Ink TUI started buckling whenever the work got real. Multiple agents in parallel, big diffs streaming in, long tool-call chains. The frontend, not the model, was the bottleneck. It was frustrating to me.
So I rewrote it from scratch in Rust on Ratatui. It talks to Anthropic's official Agent SDK through a small TypeScript bridge, so it should follow Anthropic's ToS. Tool calls, file edits, and permissions all work the same. Only the terminal layer changed.
The thing that surprised me: you don't realize how much the UI was shaping your perception of the model until it's gone. With smooth rendering and no input lag, the same Claude that felt slow and buggy before suddenly feels sharp and responsive. The model didn't change. The friction did.
It's pre-1.0. Would love feedback from anyone :)"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude","code"],"value":"Show HN: A native Rust TUI for Claude Code, because Ink got slow"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude","code"],"value":"https://www.npmjs.com/package/claude-code-rust"}},"_tags":["story","author_srothgan","story_47797324","show_hn"],"author":"srothgan","children":[47797759],"created_at":"2026-04-16T18:13:44Z","created_at_i":1776363224,"num_comments":1,"objectID":"47797324","points":1,"story_id":47797324,"story_text":"I use Claude Code every day, and the React Ink TUI started buckling whenever the work got real. Multiple agents in parallel, big diffs streaming in, long tool-call chains. The frontend, not the model, was the bottleneck. It was frustrating to me.
So I rewrote it from scratch in Rust on Ratatui. It talks to Anthropic's official Agent SDK through a small TypeScript bridge, so it should follow Anthropic's ToS. Tool calls, file edits, and permissions all work the same. Only the terminal layer changed.
The thing that surprised me: you don't realize how much the UI was shaping your perception of the model until it's gone. With smooth rendering and no input lag, the same Claude that felt slow and buggy before suddenly feels sharp and responsive. The model didn't change. The friction did.
It's pre-1.0. Would love feedback from anyone :)","title":"Show HN: A native Rust TUI for Claude Code, because Ink got slow","updated_at":"2026-04-16T18:48:01Z","url":"https://www.npmjs.com/package/claude-code-rust"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"mbadyl"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["claude","code","streaming"],"value":"I've been using Claude Code heavily and built a web interface around it. The main thing that felt missing was visibility into what the model is actually doing while it runs - so the core feature is a real-time interactive graph of Claude's steps and tool calls.
It also has a chat UI with streaming, project and session management, permission controls with YOLO mode, and profiles for switching between API keys.
Early stage, MIT licensed, building it mostly for myself."},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["claude","code"],"value":"Show HN: Clauductor \u2013 Web UI for Claude Code with real-time work graph"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/mikolajbadyl/clauductor"}},"_tags":["story","author_mbadyl","story_47328619","show_hn"],"author":"mbadyl","children":[47328789],"created_at":"2026-03-10T20:49:27Z","created_at_i":1773175767,"num_comments":1,"objectID":"47328619","points":1,"story_id":47328619,"story_text":"I've been using Claude Code heavily and built a web interface around it. The main thing that felt missing was visibility into what the model is actually doing while it runs - so the core feature is a real-time interactive graph of Claude's steps and tool calls.
It also has a chat UI with streaming, project and session management, permission controls with YOLO mode, and profiles for switching between API keys.
Early stage, MIT licensed, building it mostly for myself.","title":"Show HN: Clauductor \u2013 Web UI for Claude Code with real-time work graph","updated_at":"2026-03-13T20:15:29Z","url":"https://github.com/mikolajbadyl/clauductor"}],"hitsPerPage":20,"nbHits":373,"nbPages":19,"page":0,"params":"query=claude+code+streaming&advancedSyntax=true&analyticsTags=backend","processingTimeMS":20,"processingTimingsMS":{"_request":{"roundTrip":17},"afterFetch":{"format":{"highlighting":1,"total":1},"merge":{"mergeLoop":{"prepareNextHit":2,"total":2},"total":2},"total":2},"fetch":{"query":9,"scanning":7,"total":17},"total":21},"query":"claude code streaming","serverTimeMS":22}