Activity
Mon
Wed
Fri
Sun
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
What is this?
Less
More
349 contributions to AI Automation Society
Day 1 Newsletter build.
Day 1 done ✅ #AISChallenge What I built: a newsletter automation on the WAT framework. One workflow file, three small deterministic tools: an infographic rendered from HTML with headless Chrome, an inline-styled HTML email, and a Gmail sender. The content comes from our own study notes, so the newsletter is a weekly "what we're learning here, and what we already applied". Screenshot below. I swapped a few parts of the stack: no Perplexity (the content is our notes), images generated from code instead of an image API, and normal permissions instead of bypass. One thing I learned: WAT gave a name to something I was already doing with Claude Code skills (instructions in Markdown + scripts that do the work). What clicked: when something breaks, fix the tool and write the lesson into the workflow, not just rerun it. It happened twice today: the app password pasted with quotes and spaces made Gmail drop the connection, and the inline image showed up as a "noname" attachment. Both are fixed in the tool now, and both are written down in the workflow. One thing I'd improve: host the image instead of attaching it (Gmail still shows inline images as attachment chips), and add a small set of test inputs before letting it run on a schedule. On to Day 2 🚀
Day 1 Newsletter build.
1 like • 7h
Once it's on a schedule, a daily login check is worth five minutes. A mail connection can drop between runs without a clear error, so the first sign is often a missing send. Make the check log in and send nothing, and alert you only when the login fails.
One instruction file per project, not one for everything
My first setup had one big instruction file for the AI agent that covered everything I do. It got too long, parts of it went out of date, and in the end nobody read it. The reason is simple. A session that has everything open carries everything in its context. The agent reads rules for projects it isn't working on, and the rules that matter get lost between them. So I split it by folder. I have eleven top-level folders, and each one has its own instruction file. It loads by itself as soon as the agent touches a file in that folder. For each project, the file answers four questions: what is it about, where is the current state, what comes next, which tools exist. The top file only has pointers and the rules where breaking them was expensive. Anything about one project stays in that project. I also work with one main session per folder. I give it the folder name and it finds the state on its own. If you build for several clients or products with an agent, try this before you write the next long prompt file: one folder, one instruction file, one session.
0 likes • 7h
For corrections, each of my projects has its own git repo. When a rule changes, you edit the file and the old wording stays in the history instead of sitting next to the new one, so the agent only finds the current rule. The commit messages can double as the done log too, so the agent can see what got finished last session.
Great Week
So Today i am checking in I have finished the foundation of my EA. I also Got my first review from a website mock up I delivered yesterday. And He was super surprised on How I was able to catch the vibe and theme of the brand. I Automated my daily morning brief, weekly Prospect search and follow up. And I will post all that tommorow Day#6 and #7 because I still have some little thigs to tweaks.... but it was a awesome training .... really love it very practical.
1 like • 7h
Finishing the EA foundation and getting a review that fast is a good sign. Since the morning brief and the prospect search run on their own, check the publishing status of the Google app behind them. An app left in Testing mode drops its connection every 7 days, and that's what happened to me. I only found out when I went to use it.
Discovery done. Second 5-star from the same client
Update on the paid discovery I posted about two weeks ago. It's done. 32 hours over two weeks, closed on 9 Nov, and the same client left a second 5-star review. What he got at the end: a written plan, the MVP scope and a price for each phase. Plus what we built along the way: a site audit, competitor research, matching rules, the key screens and the tech stack decision. The line from his review I'm keeping (a marketing agency owner in New Zealand, 5-star review on Upwork): "What I've particularly appreciated is his willingness to listen, challenge ideas constructively and work through problems with me, rather than simply taking a brief and running with it." He didn't praise any code. There wasn't any yet. He praised the questions. What I took from it: 1. Discovery is a product on its own. Clients will pay for clarity before the build 2. Push back early. Challenging an idea in week 1 costs less than rebuilding it in month 2 3. Same client, two contracts. Delivery is still my best lead source The MVP plan is with him now. Has anyone here sold a paid discovery as its own contract? How did you price it?
0 likes • 7h
@Ali Abdullah Appreciate you reading it that closely, Ali. A plan the client can check line by line builds more trust than a number on its own.
0 likes • 7h
@Alexis Ampuero Good point, Alexis. The rule on its own is what gets lost when someone new picks it up. Examples with the reasoning behind them show why the exception exists.
Make an OpenAI agent run a task every morning with one API request
Want an agent to send you a status summary every morning, or follow up on something later, without n8n or an external scheduler? The OpenAI Agents API now has scheduled automations: you create one with a single request and OpenAI starts the agent session for you. It is in public beta, added on Oct 23 per the changelog. Here is how to set one up (docs: https://developers.openai.com/api/docs/guides/agents-api/sessions/automations ): 1. Create an agent with the instructions and tools the task needs, and note its agent_id. 2. Pick a target. new_session starts fresh and needs agent_id, environment and your message in target.input. existing_session keeps earlier context and needs session_id plus reminder text in target.instructions. 3. Pick a schedule. periodic takes a cron expression (a five-field pattern for when to run) plus a time zone, for example "0 10 * * *" with "America/New_York" means every day at 10 AM. once takes run_at, a Unix timestamp in seconds, set in the future. 4. Send POST /v1/agents/automations with the header OpenAI-Beta: agents=v1 (the SDKs add it for you). The doc's daily example body: 5. {"type":"periodic","name":"Daily session","expression":"0 10 * * *","timezone":"America/New_York","target":{"type":"new_session","agent_id":"agt_123","environment":{"type":"openai_hosted"},"input":[{"role":"user","content":[{"type":"input_text","text":"Summarize the current status and next steps."}]}]}} 6. Test it. The automation is saved as active but does not run right away, so trigger a manual run with POST /v1/agents/automations/{id}/invocations (no body). Check history with GET on the same path. Pause, resume and delete have their own endpoints. Gotchas: a run with status completed only means the session accepted the task, not that the agent finished. Recurring runs get a small random delay, and if the session is busy the task waits. A scheduled run also does not supply a missing approval, so check the approvals page: https://developers.openai.com/api/docs/guides/agents-api/sessions/approvals
0 likes • 1d
The gotcha that completed only means the session took the task is the one I'd build around. I'd have the scheduled task write its result to a file or a sheet with a timestamp, so the next morning you check the output itself, not the run status. The approvals note is the other one I'd check before the first run.
1-10 of 349
Tariq Osmani
6
1,100 points to level up
@tariq-osmani-1801
Founder of Smart AI Workspace. I build AI agents and n8n + Claude Code automations for B2B teams. Here to learn, share builds and meet people.

Active 1h ago
Joined Oct 26, 2026
Afghanistan
Powered by