Pouyan Salehi

And @karpathy made it even easier to apply this concept today with the idea in a gist format: x.com/karpathy/statu…

Karpathy's idea: collect raw data → LLM compiles a wiki → query against it → file outputs back so thinking compounds. I applied this to finding PMF. Every customer call, competitive signal, market article, half-formed hunch goes into raw/. An LLM compiles it into a living strategy: positioning, buyer profiles, evidence, hypotheses. Then it works for you: "Where is our positioning weakest?" answered across all signals Hypotheses tracked with real evidence for/against Health checks catch contradictions and stale assumptionsEvery query gets filed back. Thinking accumulates. We're a few days in. Already catching things we would have missed. A customer quote that contradicts our positioning, a pattern across calls we hadn't connected. The old way: call → manual summary → update docs → context resets by next week. The new way: nothing resets. Signals compound. The shift isn't "use AI to write docs." It's stop keeping strategy in your head. Feed the system, let it synthesize, decide from what it surfaces. Fast flowing water finds the path.

AI has made startups incredibly fast at shipping. A tiny team can build what used to take an entire org. But shipping speed isn't actually the advantage anymore. Everyone has access to the same tools. The real advantage is learning speed. How fast you take in signals from customers, the market and turn them into decisions. That's what finding product-market fit actually is. A learning loop. Most of us are still running that loop manually. Inspired by @karpathy's post on LLM knowledge bases, I pointed this at PMF search.

"Leverage is no longer about how much one organization can produce; it’s found in how much context people, teams, and institutions can coordinate across humans and agents.” - this is spot on, from @gabepereyra harvey.ai/blog/autonomou…

AI Ops is going to be one of the most important roles at companies. Context engineering. Agent orchestration. Debugging. Token mgmt. And so much more. The companies that are seeing meaningful impact across the org have at least one person focused on this.

Giving LLMs the right context is one of the biggest gaps I'm seeing when deploying AI in sales (for meaningful use cases and driving high adoption across a team). @JasonSCui and Jennifer Li wrote a great piece that describes this challenge in more detail. And it's aligned with @ClearskiesAi is building - Context Infrastructure for Revenue teams that want to build with AI. Exciting times ahead.

introducing 🐺 AlphaClaw, the ultimate setup harness for @openclaw. open-source, self-managed, free-to-use, with no lock-in. AlphaClaw makes OpenClaw setup and maintenance easier by providing an elegant GUI that wraps OpenClaw's CLI. 📅 Google Workspace OAuth & pubsub built in w/ gog-cli 🔄 Auto-backup to GitHub 🧱 Prompt hardening reduces agent drift 🩺 Drift Doctor analyzes your prompts and workspace for drift 💬 Telegram multi-topic workspace setup wizard 📂 Full file browser and editor, no SSH needed 🐕 Watchdog auto-detects crashes, self-heals gateway 🛠️ Manage env vars from the UI 🔑 Manage model keys & OAuth visually 🪝 Webhook creator & inspector with replay & debug 📊 Token usage & cost analytics built in ⬆️ One-click updates, no redeploy needed 📦 Import existing setup from GitHub i didn't build alphaclaw to replace openclaw or compete with it. openclaw is the best user-owned AI agent framework out there and more people should be able to use it without wrestling a CLI for two hours. there are so many managed "deploy your AI in seconds" product. but they lock you into their platform. if they pivot, shut down, or jack up pricing, your agent goes with it. alphaclaw gives you that same one-click simplicity, but everything runs on your infra with your data. no proprietary backend. no config hostage. if railway disappears tomorrow, you still have a standard openclaw instance backed up to your own github repo. everything alphaclaw does, you could do manually. it's just automation and UI on top of the real thing. outgrow it? disagree with its opinions? eject. your openclaw instance is still a standard openclaw instance. to make it convenient, i’ve created both a one-click deploy template on railway and render to start quickly. make sure you have 8GB of ram on your instance. look forward to your feedback and to building this out with the @openclaw community! 🦞 github.com/chrysb/alphacl… feature deep-dive in the 🧵

As an Iranian living inside my country during war, I will say this plainly. The chance of me being killed by masked, armed men roaming the streets, or by gunmen standing on mosque rooftops behind machine guns, is far greater than the chance of dying in an airstrike. That is the truth of life under the Islamic Regime occupying Iran. The real fear here has never come from the sky. It has always come from the men the regime unleashes among its own people. #Irán #IranIsraelWar

The biggest problem with AI for sales teams right now isn't the AI. It's the data underneath it Here's what I mean Your customer lives across 5+ systems. They're a record in Salesforce, a transcript on a Gong call, an email thread in Gmail, a calendar invite, a Slack mention. Five partial versions of the same customer, none of them connected So when you ask an agent to "prep me for my 2pm call with Acme" — it gives you a confident-sounding brief based on 20% of reality To actually fix this you need to build a customer context graph — a unified layer that gives you a complete picture of every customer across all your systems. This is what it takes - First you need to build integrations into every system. Salesforce, Gong, Google Workspace, Slack. Each has its own auth, data model, and rate limits Then you need entity resolution. "Acme Inc" in Salesforce, "Acme" in Slack, "ACME Corporation" in Gmail — same customer. Same thing at the contact level. The same person is a CRM record, a call participant, an email sender, and a Slack mention. You have to resolve all of that into one identity Then you need activity mapping. Every call, email, meeting, and Slack thread needs to be normalized and linked to the right account and the right people. Not just stored — connected Then you need real-time sync. This isn't a one-time ETL. Deals move, calls happen, emails come in. A batch job from last night is already stale Then you need to build the actual agents on top. Prompts, retrieval logic, access controls so reps only see their accounts. And deploy them into Slack or wherever your team works And then you need to maintain all of it. APIs change, schemas break, your team's process evolves. You're now running an internal data platform Skip any of these steps and your AI just hallucinates confidently with partial data. Which is worse than no AI at all Want this but don't want to build and maintain it all - use @ClearskiesAi — the customer context graph for revenue teams. CRM + calls + email + calendar + Slack. entity resolution, activity mapping — all linked and giving your AI full customer context in minutes.

This.

Imagine describing Pol Pot as “ideological electrician who made Cambodia a regional power”. That’s what this is. Khamenei killed, tortured, and imprisoned millions. He was likely the most evil man on the planet.

Make it make sense.

Iranian women are dancing in the streets—without hair coverings—after hearing that Israel killed the tyrant Khamenei. So many Western leaders should feel ashamed at their cowardly response this morning. We will all remember.

I am a woman from Iran. Let me tell you honestly how women like me see politicians like you in America, women who speak about freedom from the safety of Congress while sympathizing with our killers under Islamic regimes. You stand next to the American flag, speaking about freedom. Now you launch campaigns saying, “No War With Iran.” Now I want to introduce you to a brave Iranian woman who understood America better than you ever will. Her name is Sara Saeidi. Born and raised in Iran. 39 years old. A mother of two daughters, 19 and 6. She was shot in the head for the “crime” of peacefully protesting while wearing a sweatshirt that said MANHATTAN, with the American flag beneath it. While you stood in Congress under that flag, she carried its name on her chest, not as symbolism for a photo opportunity but as a dream. Three days after she was killed, her body was returned to her family only after threats and money were taken from them. They were banned from holding a proper funeral. Authorities falsified the circumstances of her death. That is the regime you refuse to confront clearly. She wanted the freedom Manhattan represents, the freedom to live without fear, without morality police, without a bullet in her head. The same freedom protected by the Constitution you swore to uphold. You speak of “No War With Iran,” but you refuse to condemn the war being waged against us, the Iranian people, by the Islamic Republic. More than 30,000 unarmed civilians have been killed. Women blinded. Teenagers hanged. Mothers executed. When American lawmakers like you reduce this reality to a partisan talking point, you do the regime’s work for it. Dictatorships thrive when moral clarity disappears and when lawmakers choose ambiguity instead of standing firmly with victims. You call yourself anti war. But where is your condemnation of the regime’s massacre? Where is your outrage at its war against its own people? You are anti-Iranian women.💔 Peace without justice is surrender. Your hatred of President Trump appears stronger than your love for America, stronger than your love for Manhattan, for freedom, for women’s rights. That is why you sound sympathetic to the Islamic Republic while remaining silent about women like Sara, like me, and like millions of Iranians who are victims of this barbaric regime. You celebrating hijab day in Manhattan in beautiful New York and watching women get killed in Iran for not wearing hijab. I dare you to share the picture of Sara, and say no to the war being wage by Islamist terrorist on us, Iranians.

Mr @NYCMayor, I say it as an Iranian New Yorker: You are a human garbage. When our people were getting slaughtered and hunted like animals by the Islamic Republic terrorists, you stayed silent. You didn't say a word. Now that the regime, the murderers of our people are under attack, you came out of the woodwork to defend them. You are an Islamist human garbage and I fight your agenda as long as I breathe.

To you, Zohran Mamdani! You stayed quiet when we have faced massacre, when Islamic Republic assassins were sent here in New York to kill us, stay quiet now! STOP lecturing us Iranians about peace. I don’t feel safe in New York listening to someone like you, Mamdani, who sympathizes with the regime that killed more than 30,000 unarmed Iranians in less than 24 hours. You talk about “safety” of this beautiful city? Listen carefully; Safety without justice means nothing. We Iranians do not allow you to lecture us about war while you had nothing to say when the Islamic Republic shot schoolgirls and blinded more than 10,000 innocent people in the streets. You were busy celebrating the hijab while women of my beloved country Iran were jailed and raped by Islamic Security forces for removing it. And NOW you find your voice to defend the regime? No. I will not let you claim the moral high ground. The people of Iran want to be free. Where were you when they needed solidarity? New York belongs to people who stand against terrorism not those who excuse it. This city is stronger than fear. Stronger than propaganda. Stronger than you think.

@samdblond Congrats!!

We're launching Monaco today. Monaco automates customer acquisition and revenue growth for startups. The platform disrupting sales with AI has finally arrived.

An urgent plea for help from inside Iran.

@aikarjal Exactly. Check out @ClearskiesAi - unifies data across CRM, email, calendar, calls, Slack into a customer context graph, then gives you an MCP server. Works great with Claude and @n8n_io.

Someone needs to build a company around Customer Context Graph. Collect all the threads – emails, meeting transcripts, slack messages, contracts, deliverables, detail, info, and config – from your customers into context that can be explored and queried by agents. This info is scattered between CRMs, ticketing systems, note takers, product, landing pages – it's inherently cross platform information. You need a new solution. Kind of how Segment did it trad SaaS apps. With this context, you can fire up Claude Cowork or similar for ad-hoc work or build extremely powerful agent automation flows. Expose the context as skills, MCP, and file system. Even better if you build it as open-source with a hosted option so people can take it on-prem as needed. Create a connector ecosystem around it. This will power every single next-gen AI-native full-stack business. Sort of like the context graph (@ashugarg @JayaGup10 ) that has been discussed recently but I'm thinking something very concrete: "Get me all the context about this particular customer." A customer-level, cross-system context substrate that agents can explore and act on

Chilling. 12,000 estimated dead. Listen and you can understand how.