AWS local emulator created by AI agents
Source-linked update: AWS local emulator created by AI agents
capability hub
Tool-using and multi-step task execution capability.
Timeline
Source-linked update: AWS local emulator created by AI agents
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
Source-linked update: Gvisor: The Sandbox Kernel for Agents
DeepSeek-V4.1-Flash-1017 Update DeepSeek V4.1 Flash has been upgraded to DeepSeek-V4.1-Flash-1017. The official web and APP models have also been updated to the new version. The API calling method and pricing remain unchanged — simply use deepseek-flash to access the latest model. Performance–cost Pareto improvement Compared with the previous DeepSeek V4.1, the new model delivers substantial improvements across benchmarks: HLE (w/o tools): 42.6 (44.0*) Terminal-Bench 4.0: 42.3 DeepSWE v1.1: 77.0 ProgramBench: 29.1 NL2Repo-Bench: 68.2 Automation-Bench v1.0.6: 60.5 Agents' Last Exam: 35.4 Apex Bench v1.1: 72.1 Chartography (w/tools): 83.0 BabyVision (w/tools): 92.5 ZeroBench-main (w/tools): 51.0 * Tested only on the pure-text subset of the HLE benchmark set. Terminal-Bench 4.0 was tested with the Claude Code framework, and DeepSWE with the Mini-SWE-V2 framework. ZeroBench-main was evaluated with Pass@5. While improving performance, we also reduced the model's task runtime and output length, enabling users to complete tasks at lower cost and higher speed and achieving a new performance–cost Pareto optimum across multiple harnesses. We specifically optimized for DeepSeek Harness, Ten
Source-linked update: Getting Robot-Use Agents to Run in Real Time
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
Hey HN. ripcloc exists because I ask agents to consider line cost in feedback early and often when they're doing work to help constrain unnecessary complexity and to trigger discussions about code smells that lead to bad architecture. It is fast enough that you can add it to a git commit hook. It is fast enough that your agent harness can run it after every tool call. The other reason it exists is to help _count your tests_. Ripcloc attributes each line as either a source line, a test line or a comment/doc line, so that you can have a better idea quickly how much your codebase is _implementation_ vs _tests_. It does this even for inline tests like `#[cfg(test)] mod tests { ... }` in Rust (and is still faster than its predecessors!).
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
Source-linked update: Show HN: Emx, an editor for co-working with agents with Emacs key-bindings
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
I built Klistra: describe your video idea and add your logo, photos or other assets, Klistra designs and animates a video for you, exported for every screen size, with fonts, media and logos kept exact. Deterministic renders make features exact and isolated, allowing for superior control and iteration compared to diffusion models. Here's a short demo render where I made an ad for Klistra itself: https://app.klistra.ai/s/4EQoWCe37SkYz5nThgegzw As a solo dev with a lot of projects going on, I found myself needing to make short promos or explainer videos. Not being familiar with animation software, I had the idea to use Claude to code up those videos and export them as video files. In the process of creating these clips I stumbled into a bunch of bugs and setup needed to make it work reliably. The list of features I wanted for better control and feedback kept growing and I needed to sign up for a bunch of services to get music, voiceovers and other media into my project. I couldn't find an existing service that had everything I wanted. Seeing these hurdles and that the technology was only easily accessible to people already confident with coding agents, I started working on Klistra, a
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
Source-linked update: Maestri – Infinite canvas for orchestrating agents
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
I have been seeing tons of stuff about people developing 'second brains' as a data vault for their AI usage on X. I fully understand the ideology, in terms of storing data, reducing context, having a single pane to work off of, etc. The part I struggle with, is that as an employee, I feel like every time I create a workflow, recurring data-pull, etc., it feels as if I am creating an app that I have to maintain myself. When my organization is constantly changing permissions, MCP accessibility, and token usage, I am having to maintain the app as if I were a developer, updating the automations as time goes on. I am grateful that I am technical enough to do this, and could probably avoid much of this problem if I had better technical foresight, but for a non-technical employee this seems like a significant barrier to AI-success on a per-employee basis. Anybody else experience the same maintenance struggle? I have developed countless useful agents that just die off in my mind after a few days because I forget about them, and they break.
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
Source-linked update: ReasoningBank: Enabling agents to learn from experience (April 2026)
DeepSeek-V4.1-Flash-1017 Update DeepSeek V4.1 Flash has been upgraded to DeepSeek-V4.1-Flash-1017. The official web and APP models have also been updated to the new version. The API calling method and pricing remain unchanged — simply use deepseek-flash to access the latest model. Performance–cost Pareto improvement Compared with the previous DeepSeek V4.1, the new model delivers substantial improvements across benchmarks: HLE (w/o tools): 42.6 (44.0*) Terminal-Bench 4.0: 42.3 DeepSWE v1.1: 77.0 ProgramBench: 29.1 NL2Repo-Bench: 68.2 Automation-Bench v1.0.6: 60.5 Agents' Last Exam: 35.4 Apex Bench v1.1: 72.1 Chartography (w/tools): 83.0 BabyVision (w/tools): 92.5 ZeroBench-main (w/tools): 51.0 * Tested only on the pure-text subset of the HLE benchmark set. Terminal-Bench 4.0 was tested with the Claude Code framework, and DeepSWE with the Mini-SWE-V2 framework. ZeroBench-main was evaluated with Pass@5. While improving performance, we also reduced the model's task runtime and output length, enabling users to complete tasks at lower cost and higher speed and achieving a new performance–cost Pareto optimum across multiple harnesses. We specifically optimized for DeepSeek Harness, Ten
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
What if every Erlang process were an operating-system process? With as much system code and drivers written in Erlang itself. This is something I was pondering for a while, and finally was able to prototype thanks to AI coding agents. erXos is a capability-based microkernel OS in Erlang: drivers, filesystems, and the network stack are all supervised Erlang processes, same as any app. The kernel itself is Fiasco, in C++, slightly modified for erXos needs. This is an OS I wanted for a long time: an Erlang machine where every BEAM process is a process visible to the OS and scheduler, not something bolted on top. You can try it right now with nothing to install: there is an emulator on the website that boots the whole OS in the browser ARM64 emulator. At the rc prompt, try `ps`, `top`, `ping` or `erl` and write any kind of erlang code. There's a Pi 4 SD image and a QEMU image too. The goal is to build an ultra-reliable microkernel OS with hot code loading and both system and BEAM-based isolation. The latter is one thing that would allow you to run millions of processes on the system, something that would be prohibitive on a microkernel alone due to the context switch overhead. Capabili
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
Source-linked update: Agents go off task for the same reasons people do
Source-linked update: OSS Minymaton (event-driven MCP message bus for AI agents, bundled SQLite)
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
I made a site for sharing small tools made with AI. With agents, it is now quick to build a tool that fits your own task. But those tools usually get thrown away after use, and the next person who may needs something similar start again from zero. So I thought maybe if result or process could shared by text or code and this can lead to kind of token sharing economy. Its name is Disposable and compared to "Open source", I named its philosophy to "Open Disposable". So the principle is "Like open source, without the upkeep. Keep the know-how, rebuild the tool your way". You can share anything you actually used and found useful, with no promise to maintain it. Others can use it as is, or give the spec that comes with it to theur own agent and have it rebuilt. If you have your own skill, prompt, or program you would like to show others, please share it. If you get stuck while sharing, let me know. Repo: https://github.com/x-ify/disposable
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
DeepSeek-V4.1-Flash-1017 Update DeepSeek V4.1 Flash has been upgraded to DeepSeek-V4.1-Flash-1017. The official web and APP models have also been updated to the new version. The API calling method and pricing remain unchanged — simply use deepseek-flash to access the latest model. Performance–cost Pareto improvement Compared with the previous DeepSeek V4.1, the new model delivers substantial improvements across benchmarks: HLE (w/o tools): 42.6 (44.0*) Terminal-Bench 4.0: 42.3 DeepSWE v1.1: 77.0 ProgramBench: 29.1 NL2Repo-Bench: 68.2 Automation-Bench v1.0.6: 60.5 Agents' Last Exam: 35.4 Apex Bench v1.1: 72.1 Chartography (w/tools): 83.0 BabyVision (w/tools): 92.5 ZeroBench-main (w/tools): 51.0 * Tested only on the pure-text subset of the HLE benchmark set. Terminal-Bench 4.0 was tested with the Claude Code framework, and DeepSWE with the Mini-SWE-V2 framework. ZeroBench-main was evaluated with Pass@5. While improving performance, we also reduced the model's task runtime and output length, enabling users to complete tasks at lower cost and higher speed and achieving a new performance–cost Pareto optimum across multiple harnesses. We specifically optimized for DeepSeek Harness, Ten
Source-linked update: Who made the strawberry? 14 days of logs building a procedural fruit with agents
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
TLDR: I built a terminal (harness) for Claude Code and Codex that works the official way, on your own subs. Anthropic stopped Claude subs working in 3P harnesses in Feb, 9 months of suffering later, I decided to build one that plays by their rules, a terminal (harness) that runs Claude Code and Codex (and others), together. (I am not claiming that this is a Deepseek harness killer lol, but it works really well for me) I use a number of agents in tandem, on my laptop and my server. Checking them from my phone took 4+ apps and I was losing hours away from my laptop, and copying context between them ate the rest. Terminess puts it all in one place. It complies with all their terms. Terminess runs the real claude and codex CLIs, unmodified, signed in with your own plans, exactly as your reg mac terminal or iTerm would. It doesn’t read, store or reuse your login tokens, and never proxies your requests, so nothing about it works around Anthropic's or OpenAI's terms (currently). Underneath is its own terminal engine, written from scratch for agent work. - Sessions live on your Mac or your own server, so agents keep going with the lid shut - You can start on your phone and pick it up on yo
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
Source-linked update: External Sandboxes for Coding Agents
Source-linked update: Standard Support: How AI agents ask for support
Source-linked update: Open WebUI – Open Terminal: A Computer for Your AI Agents
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
Source-linked update: Linux sandboxes from S3. 93-99% cheaper than always on for AI Agents.
DeepSeek-V4-Flash Update The official release of the DeepSeek-V4-Flash API is now in public beta. The API calling method remains unchanged — simply set the model name to deepseek-v4-flash to use the latest version. Significantly enhanced agent capabilities, with benchmark results far exceeding V4-Pro-Preview: Terminal Bench 2.1: 82.7 NL2Repo: 54.2 Cybergym: 76.7 DeepSWE: 54.4 Toolathlon verified: 70.3 Agent Last Exam: 25.2 Automation Bench (Public): 25.1 DSBench-FullStack: 68.7 DSBench-Hard: 59.6 Note 1: For the Code Agent tasks in the public benchmark sets, the official DeepSeek-V4-Flash was tested using the DeepSeek Harness minimal mode (to be released soon) as the framework, with the max effort level, topp=0.95, and temperature=1.0 Note 2: DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set The official V4-Flash natively supports the Responses API format and is specifically adapted for Codex. For the specific configuration, please refer to the documentation . DeepSeek-V4-Flash-0731 keeps the same model architecture and size as DeepSeek-V4-Flash-Preview, and was only re-post-trained. Note: This upd
APIDir Delta · controlled rollout
Confirm your address to join the waitlist. Editorial delivery starts only after sender, suppression, and evidence checks are active.