I've tried OpenAI, Claude, Dots, Hermes, OpenClaw, and sooo many trading strategies. I'm hoping i can help you save time and money learning from my mistakes. Ask me questions about what works, what fails, and let's develop a community together. Jesus is King :)
If the Ark discovery in Turkey is in fact authenticated, then it will be the loudest wakeup call God has ever sounded on earth as to its final days.
"But as the days of Noah were, so shall also the coming of the Son of man be."Mat24:37
Faster steering will help, but I'm still having to tell my Dot, LittleTibo, to start, continue and verify the work. It stops after partial progress and asks questions we've already answered. The first screenshot is its own summary of those failures.
I've had the same frustration with Codex Cloud: my computer is on, I've given it full access, and it asks before looking for the context. Astra feels much better locally. The second screenshot is what I reported yesterday.
Why have Opus rethink where every support message belongs? Have it help build the sorting rules, then let Jev run them.
@vesko_st just shared a skill for this. The result is a file you can read and edit when your business changes.
https://t.co/clsBJP6k9e
@biblegateway "Besides me there is no savior." I want the people who know me to know Jesus. He died for our sins and rose again, and we cannot save ourselves.
I keep asking my agents to keep working when something goes wrong. @odysseyml says Odyssey-3 did that with a robot: it adjusted after a missed grab and retrieved a dropped object, even though those recoveries weren't in the robot demonstrations.
https://t.co/W12xfb5YlY
Today we're launching Odyssey-3, the most powerful foundation world model yet.
It sets a new state of the art on Physics-IQ, and powers robots, trains AIs, and generates interactive experiences. It's really cool.
Experience the model today, all for free!
@leonabboud I'd have it return the original recording date and timestamp with each clip, and flag any prices, model recommendations or setup instructions that need checking before they go back out.
@ShaneMac@bot The task needs to save the destination's platform, workspace and channel ID when it's created. "Exec team" is a display name. If that saved destination disappears, finding another chat with the same name shouldn't count as recovery.
@Mayhem4Markets For a small business, "fix this identity permission first" is much more useful than another list of 200 vulnerabilities. Show which accounts lose access and which workflows depend on that permission before recommending the change.
@CoinMarketCap For learning, I'd pair this with the same chart cut off on September 4. Have the agent mark the range and explain what would invalidate it before showing the breakout, so the later candles can't influence its answer.
If I'm building an agent to check 80 invoices, I don't need 80 raw tool responses filling its context. Have code filter the records, then give the model the overdue accounts and source IDs.
@claudeai supports this through programmatic tool calling. Your app still executes the custom tools.
https://t.co/y2i1wls5sn
An agent can send your customer's booking confirmation, time out before recording success, then send it again on retry.
@inngest's docs explain the distinction: deduplicating the booking event doesn't prevent that. The email send needs its own stable idempotency key, supported by the email provider.
https://t.co/ULxQNjqun5
@ai_explorer25 This matters for how-to videos. If you're teaching someone to tie a knot or operate a tool, a convincing-looking motion can still be wrong. I'd use real footage for the physical steps and AI for captions, narration and layout.
I believe Jesus is coming soon. Hebrews 10:24–25 tells us to keep gathering, encourage one another and stir one another up to love and good works as we see the Day drawing near.
@nickbaumann_@ChatGPT This would be useful for comparing AI costs: let me change the input/output mix and cache rate, then watch the bill change. Keep the billing assumptions beside the result, like the payroll definition under your numbers here.
Here's a specific example of the harness getting it wrong: in LangGraph, you could answer a question, the task could finish, and get_state() would still report that question as pending while another task waited.
@LangChain fixed this in 1.2.13.
https://t.co/xGNykmSkyl
ASUS is putting Hermes, MuseTree and ComfyUI into the same local creative setup. I'd use a stack like this for repeated product-image work, with a cloud model reserved for the decisions the local model can't handle. @ASUS@NousResearch
https://t.co/pRIHHJC7Qw
Advancing our mission to deliver ubiquitous powerful agents to everyone, Hermes Agent is on the new ASUS ProArt RTX Spark PCs: MuseTree and ComfyUI integrations plus local model support on up to 128 GB of unified memory.
A complete creative stack that runs on your own machine.
@NousResearch The ASUS page says client files stay 100% local, then describes bursting to cloud models. Is that an explicit choice for each job, or can the agent switch on its own?
@robertwiblin Does the 50PB include more than readable agent transcripts? Your screenshot says data from training and testing across thousands of GPUs. The 66-million-year reading comparison seems to assume all of that is text.
@profntwright Acts 8 gives us that connection in an actual conversation: the Ethiopian asks who Isaiah is describing, and Philip begins with that very passage to tell him the good news about Jesus. Isaiah belongs in our conversations about the gospel. Acts 8:30–35.