I’ve heard lots of loathing that Seth will always cast himself in his TV shows. So let’s assume he’s going to do it. What role could he take that you’d be happy with. My personal vote is Quasar.
I’m trying to connect my Nanit to my Google home. When I finish logging in though, it just takes me to the Home Screen. I’m wondering if the addition/mandatory 2 factor authentication broke their OAuth callback? (Sorry for the tech jargon. Hoping an employee sees the likely cause)
Is anyone else currently able to make a new Google home connection?
I'll open this thread up to Dart Function feedback in general, but a particular feature request came in that I think takes some special care.
While I generally recommend params as the way to parameterize code between environments, and it is the only safe place to store API keys, I do recognize that Dart defines have advantages such as allowing dead code pruning.
My question, however, is how to best support it. We currently support environment variables and secrets using .env and .secret. This file-based convention naturally extends to multi-project support with .env[.project or alias] and even .env.local for the emulator.
Would a similar dart_defines[.project or alias or "local"].json be good for this community? If not, what else would be a better developer experience?
There’s so many possible foils left that I’m not sure how they could possibly resolve in 2 books:
Existing foes:
The Maestro
Princess Chandra
Agatha
The harem
The shadow mimics
Probable sequels:
Princess Formidable
Xuanshin Jinx’s twin daughter (SP?)
Nurse Yugoslav (I’m convinced she’s the woman on the 16th floor)
The sand ooze
On my watchlist:
Brittany/Ysalte
Odette/Nekhebet (I have a whole theory why she’s a part of the eulogist metaphor, which combined with the sins of her past means she had to die)
Eris
Samantha (once she gets her body and memories)
Apito
The war mages
OIs
Florin/his gun
Long shot concerns
Diiiiiiiiwata
Harpocrates
Either Tyrannus or Helic when they’re mad about Emberus
The brothers/mother of Sheol
How could this possibly get tied up? Did I miss any? After book 7 I was worried about Brittany, Li Na, and Emberus and at least two got taken care of + harbinger + folded Krakaren into Apito
I'm still getting oriented and want to set up my machine to help with multiple things. I have many roles and I'd like help vetting them before I set it up and have no idea why something is going wrong.
I'm a software developer and have a major home remodel to plan. Some of my goals are:
Automated MCVE, RCA, and fixes for bug reports in my OSS repos
Automated triage and fixing more trivial feature request in my OSS repos
Creating small internal coding projects (e.g. dashboards)
Crawling the internet to create daily dossiers of relevant news and sentiment for my projects as well as immediate notifications if social media or GitHub suggests an outage
Find my vision and create a parts list/project sheet for the remodel.
Get notified of any sales on my home remodel wishlist or substantially similar products
Automate system health and security
My machine is a Mac Mini M4 Pro with 48GB RAM. Here are my proposed roles:
| Agent | Role | Model | Security |
|---|---|---|---|
| Mordecai | orchetsrator/manager | Gemma4 26B | Access only to channels & agents other than Drakea (admin) |
| Carl | lead coder | Qwen3.6 35B Q5 | read & write to local project dirs. No production keys. PR create access to a GitHub organization (own identity) |
| Li Na | QA. Code reviewer & compliance (always critiquing Carl's plans) | Qwen 2.5 Coder 7B | Read-only access to Carl's output. Write access to GitHub PR comments (own identity) |
| Prepotente | Lead architect | Qwen 3.6 35B | Read access to Carl's project files. Write access to architecture MD files. Wipes the GCP repro project Samantha creates |
| Samantha | MCVE analyst ready to take a beating | Gemma4 E4B | Ephemeral docker container with a local Firebase emulator and scoped GCP project |
| Drakea | SRE (finds "traps") | Gemma4 E2B | Read-only access to all files. Reviews logs. Searches CVEs and compares to project files. |
| Zev | Social Media Manager | Gemma4 31B Q5 | Read-only API access to web search, GitHub, and social feeds. |
| Katia | Architect visionary (artist critique) | Gemma4 26B | Read-only web access for sourcing material. Write access to rendering pipelin |
| Princess Donut (finally!) | Aesthetic Director | Llama 3.1 8B instruct | Internal processing only. Communicate with Katia to dictate overarching style and vibes |
| The System AI (AKA Daddy) | Constraint Enforcer | Gemma4 EB4 | Readonly access to Katia & Donut's outputs. Enforces physics, zoning, and budgeting constraints ("we all have our limitations") |
Am I doing something stupid or is OpenClaw just slow?
I've got a 48GB Mac Mini M4 Pro. I have a TB5 SSD with my ollama models on a Samsung 990 pro. I'm trying to run a 6B quantized Gemma4 31B with 32K context. Yet it takes multiple minutes to just respond to "hello" over telegram.
Looking at the math, this _should_ fit in memory and even when I switch to a smaller Gemma model and a 16K context window, I'm still getting multi-minute responses.
Using ps I don't see any CPU contention. I've been using Gemini CLI and Gemma on OpenClaw to debug. They both say I should surely be seeing fast performance now, but it's still slow as can be.