Shubham Saboo

I just sat down with @Tiny_Fish's founder to talk about the future of web agents. 5 things from that conversation that stuck with me: 1. Agents are becoming the primary users of the web. Sites that block them today will be making themselves agent-friendly, even agent-certified, to get that traffic. 2. Human-in-the-loop is a transition phase. Next come headless agents that handle login, pricing, and multi-step execution across services on their own, with a person stepping in only for something new or risky. 3. Subscriptions don't fit agents. Tinyfish replaced them with a prepaid wallet so each agent task can be a micro-transaction. 4. Trust is the biggest competition in this space. Tinyfish keeps credentials in a vault the model never sees, and a separate safeguard model checks every action for risk before it executes. 5. Shuhao runs 12 parallel Claude Code sessions as a founder and Chief Product Officer to quickly build and iterate on new features. https://youtu.be/vCHe1CSEYvg

wtf...there's now writing weights benchmark for models. tbh i wasn't ready for this. But great to see Gemini 4 Argon doing so well...just wait for Gemini 4 Carbon!!

Google cloud revenue go brr. our 1P model APIs now process nearly 32B tokens/min

wtf is an agent harness. An agent is a model plus a harness. Model reads the context and picks the next action. The harness is all the code that turns that choice into work. Claude Code, Codex, Antigravity, OpenClaw and Hermes Agent are all harnesses. 8 common parts every agent harness shares: 1. The loop • Sends the context to the model and runs whatever tools it asks for • Feeds the results back in for the next step • Decides when the task is finished 2. Tools and sandbox • Give the model a terminal, files, a browser and your connected apps • Run every task inside its own isolated computer • Keep the damage from a bad command inside the sandbox 3. Subagents • Take side tasks off the main agent and run them in parallel • Start each one with a fresh context window • Send back only a short summary, so the main context stays clean 4. Context management • Decides what the model sees on every turn • Loads skills and files only when the task needs them • Summarizes old history before the context window fills up 5. Memory • Stores progress notes, git history and skills on disk • Lets a new session read what the last one did • Picks the work up where it stopped 6. Guardrails • Hold risky actions like emails, payments and deletes for your approval • Keep real credentials outside the sandbox • Give a prompt injection no credentials to steal 7. Verification • Stops the agent from handing you a summary that only sounds finished • Runs tests, browser checks or a separate grader first • Lets the agent call the task done only after those checks pass 8. Observability • Traces every model call and tool call • Spots the agent repeating the same failed edit • Pushes it to try a different approach follow @Saboo_Shubham_ for more on Agent Harness.

Crisp, concise and clear writing is such a huge moat when AI does all the writing.