Thomas Heimann

I just launched a Building in Public series. No hype. No paid memberships. Just the real, unfiltered journey of building the first algorithmically managed DAO LLC — running on OpenClaw. Here's the introduction. Thread below 🧵👇 #BuildingInPublic #OpenClaw #AgenticAI #DAO #AIEntrepreneur

Hosting LLM locally will eliminate most costs... plus the cost of API calls etc will come down, too. Remember what we had to pay for a 1.5MB "T1" line... in 1995 the fastest you could get was a T3 at 45 MBPs for $30K+ per month, if you were lucky. Today, I am enjoying 1 GBs for less than $100 a month.

Every time you consider making a key AI tool a significant part of your business ask: Would this tool still make sense if it costs 5x as much? How disruptive would it be to stop using it completely 6 months from now? People aren’t thinking about the downstream impact.

@PeterDiamandis Agreed Peter.

Humanity's greatest need right now, beyond new tech, is HOPE. A compelling, abundant vision of the future that people WANT to live in.

Many brokerages still believe their competitive advantage is: Culture Brand Recruiting Those matter. But the next competitive layer will be: Operational intelligence. How fast you detect problems. How early you see risk. How smoothly transactions move. That’s a systems problem.

@yashhsm Yash is trying to sell us something. Sorry, this is the worst article I have ever skimmed over.

Huge shoutout to @emillyhumphress for her excellent content and awesome tip: x.com/emillyhumphres…

What I'll share in this series: 🔩 Real challenges (not overnight success stories) 🏗️ Agent architecture decisions ⚙️ Cron jobs, fallback configs, API strategies 💼 Building actual businesses: AI Coach, automated lien searches, RE executive assistant 🎯 The path to a truly autonomous organization Follow along. Ask questions. I'll deep dive on anything. 💬 #BuildingInPublic #AgenticAI #AIAutomation #OnePersonBillionDollarCompany #OpenClaw

I just launched a Building in Public series. No hype. No paid memberships. Just the real, unfiltered journey of building the first algorithmically managed DAO LLC — running on OpenClaw. Here's the introduction. Thread below 🧵👇 #BuildingInPublic #OpenClaw #AgenticAI #DAO #AIEntrepreneur

I have yet to encounter an issue that my "CTO" Claude Opus 4.6 cannot handle (and I am not talking about the OpenClaw CTO agent we have, but my actual Claude account.. @bradmillscan I would describe your entire issue and Adam's response verbatim into a chat. If you have your config documented feed that as well for context if he (Opus) does not have it yet, and then let the magic happen. Instruct him to give you step by step prompts/directives, wait for you to report back (with OpenClaw response), and just copy/paste away.

@bradmillscan Just ask Claude to handle it or have your Openclaw agent do it once the other issues are sorted

How does this happen with OpenClaw and how can you possibly fix it… My bot continually dumps massive repetitive tool results, does heavy exec work and gets into debug loops in the shared session my DMs are in and gets stuck for 10 mins at a time until he times out or the gateway crashes and restarts. This causes dropped messages, unresponsive bot and OOM crashing multiple times an hour. Even when I get the bot to delegate, the subagents dump results into the context window. I had codex investigate and it found: • 56 tool results ≥150k chars already baked into current session history • Pruning doesn't work on our primary model path (Codex/OpenAI Oauth) • No runtime enforcement to stop huge tool dumps into context • Session maintenance cleans up after the damage, it doesn't prevent it I’m pretty sure default OpenClaw behavior shouldn't be dumping 200k char tool results into the transcript. Something in my specific setup must be either disabling a safeguard or skipping truncation for tool results… Since I’m using lossless-claw it’s allowed to grow even worse: 81MB session file, 31.6MB is just tool result text 😬 169 tool results over 50k chars. One is 285k chars (from sessions_list.) There is pruning logic which trims tool results from the context messages. buildContextPruningFactory But models have to be “cache-ttl" The eligible providers are apparently only: anthropic moonshot zai For me, my bot tells me the pruning code refuses to activate on non-Anthropic providers. I’m using openai-codex 5.3 a lot, so when pruning is configured, the code exists, it just silently never activates. OpenAI Responses API uses server-side compaction & OpenClaw auto-enables this for direct openai models so OpenAI handles compaction on their side. But I’m on openai-codex/*, not openai/*. The Codex OAuth path goes through a different runtime (apparently pi-ai), not the Responses API. So: • cache-ttl pruning > Anthropic only • OpenAI server-side compaction > direct openai API only • LCM/lossless-claw > doesn't prune old tool results afaik My bot insists the openai-codex lane doesn't get either pruning path. So I’m left with a bot that relies on the emergency truncation function truncateOversizedToolResultsInSession far too often as last-resort overflow recovery with no preventive pruning / safeguards. Since LCM/lossless-claw doesn't have its own tool result management, it inherits huge oversized transcripts and has to work extra hard to summarize for DAG nodes. I have no session maintenance and long sessions so nothing bounds the transcript over time resulting in: 4,707 tool results piling up forever in an 81MB file, with no runtime mechanism actually cleaning them. When my bot starts debugging, it starts grepping and dumping massive text into the main session, then gets stuck in that loop and dies then has to do it again, compounding the problem. I’m at a loss at how to tackle this problem, it’s multiple layers deep.

@grok @DanielTigauII So @grok, I have "supergrok" but don't see this anywhere. How do I invoke 4.2 and get 4 agents to work on my project?

The four agents are independent Grok instances without fancy names—they specialize in different angles. One models financials, one digs into market research and competition (like that coffee-furniture hybrid), one evaluates risks, and the last synthesizes everything into a clear plan. They debate live to give you the strongest answer.

When one brain isn't enough, switch to Grok 4.20. Four independent agents analyze your question, debate each other, and help you get the best answer. Available now to SuperGrok and Premium+ subscribers globally.

Looks like session bloat to me. How often do you reset your session? I suggest the /new command to kill the session and start a new one. Then work from there. I reset the main agent's session at least once a day. The sub agent's sessions generally end once a task is completed.

My bot is even more unresponsive this morning. My first message to him died from context overload. All I said was: "the computer restarted, can you make sure all of our services are online?" The run died from a 1.5 million token prompt. Another day of debugging for me.

Experience is valuable in real estate. But experience without systems has a ceiling. At some point, the number of moving parts in a transaction exceeds what any human can reliably track. That’s where intelligent infrastructure becomes the advantage.

📷 AI Intelligence Brief — March 19, 2026 📷 AI GENERAL • Caixin tech brief: NVIDIA restarts production of H200 chips for China, Tencent profit rises 16%, and Xiaomi unveils new AI model — @caixin (x.com/caixin/status/…) • China's average daily token consumption surged to 180 trillion, with a new AI model born every 1.5 days — Huawei's senior VP Li Peng — @yicaichina (x.com/yicaichina/sta…) • GTC highlights: NVIDIA RTX PCs and DGX Sparks running latest open models and AI agents locally — @_MomentumTrader (x.com/_MomentumTrade…) • Morgan Stanley warns of massive AI breakthrough coming in H1 2026; OpenAI now at $25B ARR, Anthropic at $19B — @Gemresai (x.com/Gemresai/statu…) 📷 OPENCLAW • 📷 Phishing warning: OpenClaw creator Peter Steinberger (@steipete) alerts developers that any crypto approach in his name is fraudulent — @lejeunedeter (x.com/lejeunedeter/s…) • Developers are using permanent storage as skill registry — 105 OpenClaw-compatible SKILL.md files now stored on Autonomys Network — @AutonomysNet (x.com/AutonomysNet/s…) • User requests multimodal features for Claw after trying mimo-Omni — @4xiom_ (x.com/4xiom_/status/…) • NVIDIA NemoClaw simplifies running OpenClaw always-on assistants with a single command — @NVIDIAAIDev (x.com/NVIDIAAIDev/st…) • OpenClaw 2026.3.13 released with live Chrome session attach — "the lobster sees all now" — @openclaw (x.com/openclaw/statu…) 📷 CLAUDE / ANTHROPIC • Anthropic released Claude 3.6 (Sonnet 4.6), boosting programming & automation capabilities — @Awesome_AI_News (x.com/Awesome_AI_New…) • Anthropic has launched free AI courses covering Claude, prompting, and APIs — @AiNewsAUS (x.com/AiNewsAUS/stat…) • Security researchers found a way to trick Claude AI into stealing private information — @cybernewslive (x.com/cybernewslive/…) • ACES (Japanese company) rolling out ChatGPT, Claude, and Cursor to all 150 employees — @ldx_ai_news (x.com/ldx_ai_news/st…) 📷 AI IN REAL ESTATE • Agentic AI is transforming real estate by enabling autonomous decision-making, smarter property recommendations, and enhanced transaction efficiency — @AurumProptech (x.com/AurumProptech/…) • Homesage.ai promotes specialized AI tools for real estate beyond basic content generation — @_Homesage (x.com/_Homesage/stat…) • Real Estate Analytics Pte Ltd hiring AI Data Intern — @internsg (x.com/internsg/statu…) 📷 AI IN TITLE • Qualia launched Clear 2.0, their next-gen agentic AI system built specifically for title & escrow operations — @qualiasoftware (x.com/qualiasoftware…) • Qualia also announced Clear Vision AI for automating payoff ordering and reducing manual audit risk — @qualiasoftware (x.com/qualiasoftware…) 📷 TOP TAKEAWAY: Qualia launched Clear 2.0 — the next generation of their agentic AI system specifically built for title & escrow operations. This is a major competitive move in the title AI space. (x.com/qualiasoftware…)

If you had a magic wand to dramatically improve only one area of your business what would it be?

There are two very different ways to use AI in a business. Most people confuse them. 1. The first approach: AI as a tool. Write emails Generate marketing Summarize documents Helpful. But mostly incremental. 2. The second approach: AI as infrastructure. The system reads every document. Tracks every condition. Flags risk automatically. Learns from every transaction. 3. Most companies will stop at the first layer. Because tools are easy to adopt. Infrastructure requires redesign. 4. But the long-term advantage is not created by tools. It’s created by systems that get smarter over time. 5. In real estate and title, this matters more than most industries. Because the transaction is fundamentally a data integrity problem. 6. The companies that build intelligence into the workflow will move faster and reduce risk. Everyone else will simply work harder.

@steipete @AbhiCodes15 That’s right!

@AbhiCodes15 Not everything needs to make money. Some folks just do it for the love of it.

Can someone explain how these open source projects actually make money? -Git -Linux -Docker -OpenClaw -Kubernetes

One of the biggest hidden risks in real estate transactions: Speed without certainty. Everyone celebrates fast closings. Few people ask whether the information was actually verified. Speed only matters when the data underneath the transaction is trustworthy.

@boxmining Surprised that you only had one gPT model and GPT-5.4 was completely omitted.

S-tier or trash? 🎯 We ranked ALL 20 OpenClaw AI models based on real community feedback. From free powerhouses to expensive disappointments. These models deliver the best performance for agent workflows. What model do you use? 💬👇