Activity
Mon
Wed
Fri
Sun
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
What is this?
Less
More

Owned by Max

AI threatens your job? We fix that!

Grammarly, Coda, & Superhuman! (Community, Creativity & Fun)

22 contributions to 🦞AGENTIX-AI Wizardry
Superhuman Go/Coda is quietly one of the best places to build AI-first business systems
A bit on why I keep building here; for anyone wondering whether Superhuman Coda has a place in an AI-first world. Most of the AI conversation right now is dominated by a real and reasonable fear: hallucination and non-determinism. Can you trust a system that might invent a fact, or behave differently each time you run it, to actually run your business? That’s the right question to ask — and I don’t think the answer is to keep AI at arm’s length. It is to orchestrate AI so it reliably achieves a business goal: not a clever demo, but a process a manager can trust day after day. That problem has two halves, and Superhuman Coda happens to be unusually good at both: - Deterministic work belongs in formulas. CFL, buttons, and automations give you reliable, auditable, repeatable execution — the same result every time, with nothing invented — for everything that doesn’t need judgment. - Judgment belongs in inference. Native AI blocks let you place a focused model call exactly where reasoning is genuinely required — and nowhere else, so the unpredictable part stays small and reviewable. And the Superhuman Go AI engine is excellent at taking a specification of the business workflow we want to automate and generating the Tables, Formulas, and AI prompts to implement it! Most of my experiments, a set of docs I called AgentixGo (see here) are really just probes into one question: where is the optimal line between the formula and the inference? Coda lets me move that line freely without ever leaving the document. I don’t know another platform where deterministic logic and AI judgment sit this close together, in a tool a business manager can actually read. Notion has recently launched their “Custom Agents” feature that goes part of the way along this path. But it falls short, and let’s face it; the Notion formula language, and their database model, are way inferior to Coda, and not quite fit for purpose in serious process automations.
1 like • Jun 19
as for pitfalls building workflows in coda or notion: the biggest is when makers do not know database modeling. building third normal form data models is ESSENTIAL. but most people outside of computer science have not learned how to do this. almost all the problems i get called in to fix with business automations are due to non normalized data models. makers dont know how many tables to use or how to connect them together to model the business process accurately
1 like • Aug 14
@Gt Teag what a GREAT question. there is a glass walll (invisible to newbies) that blocks makers from progressing beyond the simplest cases. and that wall is "data design". (ok. the initiated also know it, in their secret language, as 3NF). sometimes we wait until they crash into that wall and have no idea how to proceed. then we walk on water and fix the issue by drawing a silly little squiggly diagram with blobbies for objects and arrows for relationships. explaining that two headded arrows are the sign of invisible missing thingies. etc. we dont use thechie-speak if we can avoid it. but sometimes we dont wait for the crash and burn to happen. instead we tell them stories of others we have helped, and how they were saved by learning to draw squigglie blobby-charts BEFORE they dig the foundations and pour the concrete! like the admin who built an amazing system for their hiring process. everybody LOVED it. it worked for months. until one person applied for two jobs and they couldnt figure out how to handle that. they ended up making the job picker allow many jobs to be selected. but that led to problems down stream. then a few interviewers wanted to be members of more than one panel. ouch! but the big crash was the first time the panel asked people back for second interviews! that had never happened before. how to store that without messing up the scheduling logic (for applicants AND the interviewers), not to mention the email sending process. argh! i drew a blobby-chart for them with blobs for applicants, jobs, interviewers, interviews, interview-questions, email templates, etc. then we drew simple squiggly arrows to show the links between them. (the lines had crows-feet arrow heads for one to many). double-headded arrows all over the shop! thars your problem! so between applicants and jobs there is a hidden blobby: the 'application'. one link to the applicant, one link to the job. boom. between the interviewer and the panel blobs there was a hidden blobbie: the 'membership' record. one link to the interviewer, one to the panel. boom!
Workshop 6: New Superhuman/Grammarly AI Agentic Workflow
Grammarly is being rebranded as "Superhuman Go" with new integrations with the other products of the Superhuman family. Coda.io is being rebranded as "Superhuman Docs" and features a significant upgrade to its AI capabilities. I have exploited this new AI side panel (which is based on Anthropic LLMs) to implement complex Agentic Workflows, withour needing any Packs, Formulas, Buttons, or Automations. The workflow is defined as a set of SKILLS in English that are stored as Coda Pages. The data is stored in normal Coda Tables. The AI system "executes" the SKILLS pages and uses MCP to read and update the Coda tables as it processes the workflow. This technique will be familiar to users of Claude Code, Claude Cowork, Antigravity, Cursor, Open Claw, etc. It also beats the capabilities of the recently launched Notion Custom Agents. This video demonstrates a use case where we replace experienced Civil Engineers to read the customers requirements for industrial water pumps in mines and quarries. The complex rules that govern the configuration of thes large (4 ton) water pump systems are defined in the SKILLS pages. When the configuration is completed, the AI checks results and, if needed, back-tracks to find the optimal solution automatically. In the next workshop I will be going into much more detail about how we buid these Agentic Workflows with the assistance of the AI and expert inputs from the engineers, their procedure manuals, and reference materials. I was unable to publish this sooner due to the Superhuman Beta Program NDA. But now the product has launched, I can share this here. respect, 🦞Max
2 likes • Jul 16
@Oliver Schlachta , its great to hear this. My mission is to use my own experiences to inspire ordinary business folks to embrace the use of AI Agents in their business processes. So it is always a great delight to hear that my efforts are having the intended effect!
3 likes • Jul 18
@Nina Kastenauer , thanks. I feel I must be missing something in my web crawling. To me, this approach I am taking, is THE best way to harness AI for business processes. A spectrum of techniques, from highly 'orchestrated' deterministic agents (using nocode formulas) at one end, through to totally AI driven, skills-page driven agents at the high end. But the big labs just want people "doom prompting" to infinity and beyond with zero guidance on how to build real business solutions. Even Replit and their ilk do not offer a citizen-developer friendly way to deploy *intelligent* workflows to your teams, departments, partners, and customers! So one guy in his mom's basement is left to figure it out and make how-to videos? (OK - it's a converted chateau dungeon and it's not my mom's - but thats not the point!) Notion and Coda (aka Superhuman now) have built the perfect framework for this, but they seem uninterested in promoting its power to the experienced business managers who are best placed to exploit it. A few years from now some spotty-faced billionaire will have built an empire on this approach, and I will be saying "I thought if it first" to some locals in a dingy taverna in Greece. C'est la vie!
Curious re: Superhuman's new vision for AI pricing
As some of you may have noticed, Superhuman has recently cancelled CODA AI billing, and placed Coda AI into a free beta. The understanding is that they will be introducing a new billing framework at some point in the future. I'm curious if anyone has any intel, reflections or insights into what their direction and vision might be for the future of pricing AI?
2 likes • Jul 8
My clients and I have become very price insensitive regarding Ai tools. We see such huge benefits that impact profits directly, that the cost of tokens is of no real concern. However, this comes from a serious focus on REDUCING token consumption to achieve maximal effeciency in our inferences. Not because of cost, but because Ai errors grow geometricaly with context window size. Deterministic results depend greatly upon providing clear, curated, focused, task specific inputs to the models. We break down big tasks into smaller steps and orchestrate our agents to do just one step at a time. Token bloat is easy. But the errors like hallucinations, loss of the plot, not following instructions, vague generic responses, etc, are the result. We expect providers like Superhuman to pass on costs for tokens with minimal margins, and charge maximal rates for real value added features. And then we look at the ROI to see if its worth the candle. So far it has been a great deal.
WORKSHOP 5: Ai Agents 🦞CodaClaw & 😈CodaYoda
I have uploaded the latest video into the CLASSROOM and on YouTube 🦀OPEN CLAW INNER WORKINGS→NOCODE DEMO🤖 We will build a working replica of the innermost loop of OpenClaw to illustrate how it operates. This innermost loop will take a prompt and generate an executable set of TOOL requests to implement that. The TOOLs will modify data in the document by executing Action Formula Language (CFL) commands that can add/delete/edit rows in tables, change control-values, bring up dialogs for the user to see and change, modify information stored in the document's pages, etc. This replicates the way Claude Code, Claude Coworker, OpenClaw, OpenAI Codex, Cursor, work. (What I call the 🦞Lobster-like Agentic Tools). This will build the core functionality that we will use in subsequent workshops to build a complete replica of OpenClaw or Claude Cowork that runs (safely) inside the no-code environment (but can take real-world actions via API integrations). I will be using Coda.io for these demonstrations because it has the most complete set of features needed. But the principles will be equally applicable to Notion, Excel, Airtable, and other no-code platforms.
2 likes • Apr 1
This is PART 1 of a 2-part series on this topic. The first video covers building the "innermost loop" of the Lobster-like agentic system. It just takes a task definition in English (and a description of the data scheme) and then executes the CFL action formula that will carry out the task. The second part will cover building the complete CodaClaw agent that works just like OpenClaw or Claude Code, but uses Coda Pages and Tables instead of the file-folder tree structure. This allows us to create and manage our Lobsters inside the Coda platform. (They can still take real-world actions through Coda Packs).
WORKSHOP 5 AI Agents; 🦞CodaClaw & 😈CodaYoda
I have uploaded the latest video into the CLASSROOM and on YouTube 🦀OPEN CLAW INNER WORKINGS→NOCODE DEMO🤖 We will build a working replica of the innermost loop of OpenClaw to illustrate how it operates. This innermost loop will take a prompt and generate an executable set of TOOL requests to implement that. The TOOLs will modify data in the document by executing Action Formula Language (CFL) commands that can add/delete/edit rows in tables, change control-values, bring up dialogs for the user to see and change, modify information stored in the document's pages, etc. This replicates the way Claude Code, Claude Coworker, OpenClaw, OpenAI Codex, Cursor, work. (What I call the 🦞Lobster-like Agentic Tools). This will build the core functionality that we will use in subsequent workshops to build a complete replica of OpenClaw or Claude Cowork that runs (safely) inside the no-code environment (but can take real-world actions via API integrations). I will be using Coda.io for these demonstrations because it has the most complete set of features needed. But the principles will be equally applicable to Notion, Excel, Airtable, and other no-code platforms.
1 like • Mar 26
All is well at home again, thanks for the good wishes. I have recorded the next workshop and will edit it for better brevity and focus. Then I will post it to the classroom tomorrow. I covers the inner workings of the OpenClaw-like agentic agents that are exciting the AI world right now - and outlines my techniques for recreating one in Coda (my CodaYoda and CodaClaw projects). Will update members by email when it is posted. Max
0 likes • Mar 31
WRKSHOP 5: AI Agents 🦞CodaClaw & 😈CodaYoda Uploaded I have uploaded the recording to the CLASSROOM and on YouTube
1-10 of 22
Max O'Brien
4
46 points to level up
@max-xyzor-9742
✨Building 🦞AI Agentic Business Processes 🤖 using Nocode Tools 💻(Notion, Coda, Airtable, N8N, AppSheet, etc)

Active 3h ago
Joined Sep 21, 2025
Gavdos