Boxmining v2

Tag: context window

  • OpenClaw 2.23 Update: 1 Million Token Context, Model Freedom, and More

    OpenClaw 2.23 Update: 1 Million Token Context, Model Freedom, and More

    OpenClaw just dropped version 2.23, and honestly, this one’s packed. We’re getting updates almost daily at this point, but I know you’re busy — so let me break down the features that actually matter and what they mean for your AI agent setup.

    The Big One: 1 Million Token Context Window

    Let’s start with the headline feature — a 1 million token context window, now available in beta for Claude Opus and Sonnet. This is massive. To put it in perspective, that’s roughly five times larger than previous context limits, and you could fit about a quarter of the entire Harry Potter series in a single conversation.

    Why does this matter? One of the biggest reasons your AI agent sometimes feels “dumb” is context overflow. Every conversation you have with your agent sends the entire chat history along with it. When that history exceeds the context window, the agent starts forgetting things — earlier instructions, preferences, important details. It’s not that the AI got stupider; it literally ran out of memory.

    With 1 million tokens, your agent can hold an entire day’s worth of back-and-forth without needing to restart or losing track of what you discussed that morning. If you’re using your agent daily for scheduling, research, or project management, this is a game-changer. The catch? It’s expensive. Running Opus with a full 1 million context window will cost significantly more than shorter conversations.

    Model Freedom: Save Money Without Sacrificing Quality

    This is where things get practical. OpenClaw 2.23 makes it easier than ever to switch between AI models on the fly. The idea is simple: not every task needs the most powerful (and expensive) model.

    Need to build a quick dashboard or generate a simple script? Send it to MiniMax or Kimi — they’re fast and cheap. Need deep reasoning, complex scheduling, or life-planning tasks like booking restaurant reservations? Keep that on Opus, where the extra intelligence actually matters.

    Speaking of which, Anthropic just released Claude Sonnet 4.6, which approaches Opus-level intelligence at roughly half the cost and twice the speed. It’s already available on Amazon Bedrock and GitHub Copilot. For most everyday agent tasks, Sonnet 4.6 is probably the sweet spot between performance and cost.

    Sub-Agent Orchestration: Let Your Agent Be the Boss

    Sub-agent orchestration isn’t technically new in OpenClaw, but 2.23 refines it further. The concept is straightforward: instead of you managing multiple AI agents individually, you designate one agent as the coordinator. That coordinator spins up sub-agents for parallel tasks, collects the results, and reports back to you.

    Here’s a tip that came up in our discussion: be explicit with your agent. Tell it “you are the orchestrator, you command the sub-agents, I don’t want to interact with them directly.” Early on, one of our partners was manually talking to eight different agents individually — that’s doing it the hard way. Let the coordinator handle delegation. You just talk to one agent.

    This pattern is especially powerful for research workflows. Your main agent can spin up multiple research bots simultaneously, each investigating a different angle, then synthesize everything into a single report.

    Video Understanding: Cool But Expensive

    OpenClaw now supports video understanding through Moonshot integration. Your agent can literally watch videos and process visual content. Before you get too excited though — this is still in the “cool but impractical for daily use” category.

    The computational cost of processing video is significant. We’re not at the point where your agent can binge-watch Netflix and write you a sequel. It’s more suited for specific use cases where you need an AI to analyze video content for work purposes. We’ll be doing a deeper dive on Kimi Vision in an upcoming video.

    Security Hardening and Cron Job Fixes

    On the security front, the advice remains the same: use a VPS. The way we’ve always recommended installing OpenClaw — on a separate virtual private server rather than your main machine — is still the safest approach. Don’t let your agent access every aspect of your life just yet.

    Cron jobs got some attention too. These are your scheduled tasks — daily news briefings, morning presentations, automated reports. They’ve been a bit unreliable for some users, particularly when old cron jobs pile up and try to execute simultaneously. The fix? Clean up your old cron jobs first, and consider having your agent build a dashboard to track all scheduled tasks. That way you can verify everything is running on schedule. Don’t trust, verify.

    OpenClaw’s Future: Foundation Model

    One important piece of context: OpenClaw founder Peter Steinberger announced on February 14 that he’s joining OpenAI to work on bringing agents to everyone. OpenClaw itself is transitioning to an independent open-source foundation, with OpenAI’s continued support. So despite the leadership change, the project remains open-source and actively developed — as evidenced by the rapid pace of updates we’re seeing.

    How to Update

    Updating is dead simple. Just tell your agent: “Hey, update yourself to the latest OpenClaw.” That’s it. Works about 95% of the time. The other 5%? Well, as we like to say — when you’re using AI, you’re playing casino. But we have tutorials for when things go sideways, so don’t worry.

    The Bottom Line

    OpenClaw 2.23 is a solid update. The 1 million token context window is the standout feature for power users, while model freedom is the real money-saver for everyone else. Sub-agent orchestration continues to mature, and the security and cron improvements address real pain points.

    If you’re running OpenClaw, update now. If you’re not, check out our setup guide to get started. And if you want to see more content like this, join our community over at BoxminingAI and drop a comment — our bot actually reads all of them now.

  • Why Your OpenClaw Agent Gets DUMB (Context Window Explained)

    Why Your OpenClaw Agent Gets DUMB (Context Window Explained)

    If you’ve been running an OpenClaw agent and noticed it getting progressively dumber throughout the day, you’re not alone. In this video, we break down exactly why this happens and what you can do about it. It all comes down to one thing: the context window.

    What Is the Context Window?

    Think of the context window as your AI agent’s short-term memory — its working brain. Every message you send, every file it reads, every task it processes takes up space in that window. It’s measured in tokens (roughly 4 characters per token), and every model has a hard limit.

    The best analogy is a human assistant who’s been given too many tasks at once. Tell them to handle your car, your house, your parents visiting, your dinner reservations — at some point they get overloaded and start dropping balls. That’s exactly what happens to your AI agent when the context window fills up.

    Research backs this up too. A 2025 study by Chroma Research called “Context Rot” tested 18 different LLMs and found that models do not use their context uniformly — their performance grows increasingly unreliable as input length grows. Even for simple tasks, LLMs exhibit inconsistent performance across different context lengths. The longer the context, the worse the reasoning gets, especially for multi-step problems.

    Why Your Agent Wakes Up Already Loaded

    Here’s something that surprised us. Every day, OpenClaw essentially kills your agent and restarts it fresh. It wakes up, reads its long-term memory files (your SOUL.md, MEMORY.md, AGENTS.md, and other config files), and loads all of that into the context window. It’s like an assistant coming to work, reading their briefing notes, and getting up to speed.

    The problem? If you’ve stuffed those files with your life story, your preferences, your childhood memories, and every random thought you’ve ever had — your agent wakes up with a context window that’s already half full before it’s done a single task.

    In our test, Jeff (running on MiniMax 2.5) woke up at the start of the day already at 136K tokens. That’s because in the early days, the common advice was to “blast your agent with your life story so it understands you better.” Turns out, that’s actually counterproductive. All that irrelevant context is eating into the space your agent needs for actual work.

    Cheap Models Get Hit Harder

    Not all models handle large context equally. We compared two setups side by side:

    Stark running on Claude Opus — woke up at around 100K out of 200K capacity, and still performed fluidly. Opus is genuinely good at working with large context windows and maintaining quality throughout.

    Jeff running on MiniMax 2.5 — started struggling almost immediately. As one of our viewers, Note, put it: “The moment you go above 120K context window, it feels like I’m talking to ChatGPT 3.5.”

    There’s a hidden reason for this beyond just model quality. To save costs, cheaper models like MiniMax aggressively dump parts of the context they consider unimportant. This is an internal optimization to reduce compute costs — but sometimes what they dump is actually critical to your task. You might ask it to make a presentation and halfway through it forgets what the presentation is even about.

    This aligns with what researchers have found: relevant information buried in the middle of longer contexts gets degraded considerably, and lower similarity between questions and stored context accelerates that degradation.

    How to Keep Your Agent Smart

    Based on our testing, here are the practical tips that actually work:

    1. Trim your memory files. Go through your SOUL.md, USER.md, and other long-term storage files. Remove anything that isn’t directly relevant to the tasks you need your agent to do. Your agent doesn’t need to know your life story — it needs to know how to do its job.

    2. Specialize your agent. AI models actually gravitate toward specialization. Instead of making your agent a general-purpose assistant that handles everything from dinner reservations to research reports, train it for specific tasks. In our test, Stark was trained specifically for making presentations and research — and it delivered significantly better results than Jeff, who was loaded with general life context.

    3. Monitor your context usage. You can simply ask your agent “How much context are you using?” and it’ll tell you. On the OpenClaw terminal, it sometimes displays this automatically. Keep an eye on it throughout the day.

    4. Clear context when needed. If you feel your agent getting dumber, start a new session. This kills the current context and lets the agent restart fresh. There’s also a natural compacting stage where the agent automatically summarizes and compresses older context — similar to how your own brain forgets the details of brushing your teeth but remembers the important meeting you had.

    5. Choose your model wisely. If you’re on a budget with MiniMax or other Chinese models, context management becomes even more critical. These models aggressively optimize to save compute, which means they’ll cut corners on context retention. If you can afford it, models like Claude Opus handle large context windows much more gracefully.

    The Bottom Line

    Context window management is probably the single most impactful thing you can do to improve your OpenClaw agent’s performance. It’s not about giving your agent more information — it’s about giving it the right information and keeping that working memory clean.

    The takeaway is simple: less irrelevant context equals a smarter agent. Trim the fat from your memory files, specialize your agent’s purpose, and don’t be afraid to restart sessions when things get sluggish. Your agent will thank you — by actually being useful.