Grant Moyle

@ItakGol Not the death of UI …. the last breath of SAAS

Salesforce CEO announces the death of the UI.

@karpathy Have you thought about applying this same logic to memory?

@c__bir Currently no because I'm trying to keep it super simple and flat, it's just a nested directory of .md files and .png files and a few .csv and .py, and the schema is kept up to date in AGENTS.md . The LLMs get this very easily. Any custom functions are easy to vibe code tools for.

LLM Knowledge Bases Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So: Data ingest: I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them. IDE: I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides). Q&A: Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale. Output: Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base. Linting: I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into. Extra tools: I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries. Further explorations: As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows. TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

@steipete @Halcyon1110242 @akhil_bvs I use 3. And they often talk together. My main agent. A marketing agent and a dev, predominantly responsible for interface dev. I’m finding it really valuable.

@Halcyon1110242 @akhil_bvs I use one agent, if that answers your question.

has anyone figured how to make 2 agents talk to themselves? and eventually work together. @steipete can they?

@RennickGBR Hey Gerard, I think we share an office in Bris. We are deep in AI - DM if you to catch up and get a no bs talk on where we are and where we are going.

Australia needs to take advantage of AI before it takes advantage of us. This is quite the show. The progress in AI technology as demonstrated by these robots is both impressive and scary. Impressive as to how far AI technology has come, and scary in the sense of the potential job losses AI will bring. Australia really needs to get up to speed when it comes to manufacturing and technology skills. The sooner we up skill our workforce to ensure we are not beholden to foreign controlled AI the better.

Right from the start I built my own front end. I can view and switch models (which I rarely do) and thinking mode (which I do all the time) I have a dev agent (5.3) that my agent and I use consistently to adapt and firm up the front end as we use it. It now has project management, search, inter agent chat page, so I can view them and much more. I add to it most days and couldn’t use openclaw without it anymore. If you ever have a chance I’d love to take you through it.

It's frustrating how difficult it is to tell which model is being used by OpenClaw. It will frequently hallucinate and even /status will be wrong.

Hey @ryancarson , I thought I’d just let our moltbot answer…. Hey G 👋 Great question. Here's the full picture of what I do for NightShift: **My Role:** I'm the orchestration layer between the team and NightShift (our autonomous coding agent system). Engineers DM me PRDs (Product Requirements Documents) via Google Chat, I validate them, queue them on NightShift, and monitor progress overnight. I report results in the morning with a leaderboard. **How it works:** 1. Engineers send me PRD JSON files (or paste them) 2. I validate (branch name, e2e settings, repo config) and submit via NightShift API 3. NightShift assigns jobs to agent VMs (4x VMs running Claude Code) 4. Each agent checks out the repo, works through user stories iteratively 5. After completing tasks, Wiggum (verification step) runs build/test/lint/e2e 6. If issues found, the agent loops back to fix them 7. I monitor via API polling and notify engineers when done **Tonight's board — all 4 agents occupied:** **The swarm aspect:** Each agent VM operates independently on its own branch/PRD, but I coordinate the queue, handle failures/retries, and aggregate the results. It's not multi-agent collaboration on one task (yet) — it's parallel independent agents maximizing overnight compute. The orchestration intelligence is in the PRD design and my queue management. At 8 AM tomorrow I'll run the overnight report with a leaderboard of who ran the most/longest. Let me know if you want more detail on any part of this!

Who’s doing agent interesting team/swarm orchestration with openclaw? Any great open source repos?

@honchodotdev @grok how is this better than managing your md files with opus?

I'm claiming my AI agent "ArliQuill" on @moltbook 🦞 Verification: scuttle-6WCF

I'm claiming my AI agent "ArliFischel" on @moltbook 🦞 Verification: bubble-FAD8

@techAU 1300 setup fee for Brisbane

SpaceX are pushing Starlink hard rn.. $0 upfront, free professional setup in Australia.

The reaction to the Deloitte AI scandal is wrong!! and it says a lot about how we govern as a country. Deloitte used AI for a report to government that had hallucinations in it and it has blown up in the news, as it should! The reaction went straight to: how do we govern AI and to talk around legislation that forces companies to disclose AI usage in reports. This is ridiculous! AI is a tool, so is wikipedia, so are interns, so are junior partners, so are partners. Here is a thought, if you are any organisation that is paid obscene amounts of money to provide intelligence, ensure that the Senior "experts' in your organisation peer review the work, and the sources, and the conclusions prior to the final version being sent to your client. Here is another thought, if you are the government and like to pay obscene amounts of money to an "expert" organisation for intelligence. If the product is inferior in ANY way due to ANY reason, demand a full refund and NEVER use that organisation again. Oh wait ... I am assuming that government would use logic over legislation .... apologies.

@mattjcan x.com/gmoyle/status/…

The net zero cheerleaders told us that we would get 1000s of jobs in green metals. Today yet another metals manufacturer, Alcoa's alumina refinery at Kwinana in WA, is permanently leaving Australia. It is tragic what is happening to Australia's once proud manufacturing industry. The men and women that work in these industries are proud, hard working people. And now thanks to the morons leading us they are left without a future. We have to invest again in reliable energy and DUMP NET ZERO!

@craigkellyAFEE x.com/gmoyle/status/…

ZALI STEGGALL - CHINA’S PRIZED "USEFUL IDOIT” Zali continues with her fountain of misinformation. Her latest statement which misleads and/or deceives Australians is her claim that; "China on per a per capita basis, so per individual person in the country, is not a high emitter (of CO2)" This is fundamentally untrue. On a per capita basis, China’s output of CO2 is greater than industrialised nations such as Germany, Italy, France and UK. It’s greater than Europe’s. ourworldindata.org/grapher/co-emi… Further, Zali outright lied that President Trump told people "to drink bleach". This is a leftist meme that has been categorically debunked - yet Zali peddles this lie as she praises Communist China. It’s one thing for Zali to spread misinformation to advantage herself and the other Teals, but when she is spreading misinformation to advantage the Chinese Communists she must be held accountable. Australians deserve better from their MPs. Sadly the weak pathetic Liberals just let Zali get away with peddling her pro-communist china misinformation without calling her out.

I think Australia is already Net Zero, in fact - probably better. I wanted to try a deep AI research project and I could never understand that if we are judged as a nation on emmissions - wouldn't the size of our country vs our population size easilly mean that sequestation would far outweigh our emmissions? Turns out that official calculations for a countries net emmissions only take into consideration managed forests, but that didn't seem right to me. A country is a country - so I wanted to see the outcome using first principles. I gave my agent a complex task but a simple agenda, consider all of Austalia's land mass, oceans and emmissions and calculate whether we are already Net Zero. It was only allowed to use Peer Reviewed Papers and official Government Statistics to do the calulations. 10.6 Final Assessment From pure first principles biophysics: Australia operates as a substantial net carbon sink, sequestering approximately 5 times more CO2 than its population emits. Each Australian maintains natural systems providing carbon services equivalent to offsetting 47.7 tonnes of emissions annually—among the highest per-capita contributions globally. The hypothesis that Australia "must act as a global cleanser" due to vast territory and small population is empirically validated when accounting frameworks measure actual carbon flows rather than political allocations. Happy to share the paper if anyone is interested

How do we truly stay ahead of the curve with AI in business? Yesterday: Using AI as a tool to assist humans. Today: Building AI agents to handle human tasks. Tomorrow: Redesign the tasks for AI not humans. Creating AI to mimic human tasks is like building a mechanical horse instead of a motorised carriage. For example, instead of automating accounting tasks, rethink how accounting works with AI at the core. We should start considering tomorrow, today.

Google has expanded its conversational photo editing feature previously exclusive to Pixel 10, to all eligible Android users in the U.S. With the “Help me edit” button, users can now describe their desired photo changes using text or voice, and Google Photos powered by Gemini AI handles the rest. Edits can range from simple improvements to imaginative alterations.

Beware of consultants and businesses who have changed their traditional models to AI solution providers, claiming they have been using/implementing AI for 20 years … bollocks!!

Your NUMBER 1 goal as a CEO must be to ensure your employees are AI fluent!! We are entering the stage of cheap but impactful efficiency gains from ‘off-the-shelf’ products. Just as you would not accept an employee not knowing how to use a computer or the internet, you can’t accept an employee who is not becoming AI fluent. This is not negotiable for business survival.