GemSpecter

I design AI automation systems for marketing and e-commerce, and I spend my creative time making photos, videos, and visuals with AI.

[Brad Groux @BradGroux]

I've propped up Elon in just about every talk I've given, but Grok Bot usage got cut over the weekend with no heads up to paying customers.

So, @DigitalMeld pulled our @xAI partner application, and the dozens of @Bot videos I had planned are scrapped.👇 https://t.co/nIkxTtOhXt

[Alex @alexmichael]

Grok Bot users are reporting their weekly usage suddenly spiking overnight, some going from 40% to 90% in a few minutes with nothing running

SpaceXSI opened an incident for this on their status dashboard, now marked resolved

Let’s see if @spacexai resets usage for the people who got hit

[GemSpecter @GemSpecter]

heads up if you're using Import Bot, I ran a quick test, and when you delete it, everything it saved to shared memory goes with it 😅 my other bots lost all of it. My workaround solution is to have your main bot re-save the important stuff before you delete it. @bot would be cool if those stuck around.

[GemSpecter @GemSpecter]

@jumperz Is tier list based on the highest github issues?

[GemSpecter @GemSpecter]

@singhai @larsencc @milichab @shahmeernavid_ @bot

[GemSpecter @GemSpecter]

Finally I added some guardrails on the user interaction. View and edit are allowed. Anything that needs a delete, a technical setup, or a payment, my Grok Bot stops and asks for my approval. Tested every scenario multiple times. Of course it's my gf, so I wasn't worried here lol. 😁

[GemSpecter @GemSpecter]

Yeah that tracks. My Grok Bot and I never hit it because we only use Slack, which is already on that list. We didn't build our own app. Here's the connection. It just posts from my connected account. The webhook thing only shows up if you want a platform they didn't build. We stayed on the built-in one, so we never ran into that. After testing was successful, I asked it to reply as Nebula.

[GemSpecter @GemSpecter]

I know prompting can help with this, but I feel like it would better by default. Sometimes the bot ends a response where I’m not sure if it’s asking me a question, suggesting a next step, or waiting for me to confirm something.
It’s a small thing that will make it seamless and make conversations feel a lot more proactive.

[GemSpecter @GemSpecter]

@leerob @bot Grok Chat can use X search queries, read, and send messages to Grokbot, yet Grokbot cannot do any of those. I has to automate a search query and ask Grok Chat to send it to Grokbot instead.

[GemSpecter @GemSpecter]

@ElSchumi @leerob @bot I didn't expect my bot to know how to use Midjourney, but after a few corrections, it started batching images and verifying if they met my expectations.

[GemSpecter @GemSpecter]

1. A fresh session with same bot keeping its instructions & memory while dropping the old chat context cost.

2. Wasnt aware deleting a bot also deletes the shared facts that bot wrote.

3. Usage only shows a percentage, so I can't see what bot or group actually burning the limit.

4. Settings for each grokbot.

[GemSpecter @GemSpecter]

@BigAL1486 @leerob @bot When I first got grokbot, I expected it to fail and not perform exact how I wanted. So after a certain task that didnt go well, I spoke to the CoS and give him my feedback and asked it to have a retro with all bots involved in that task. Saw improvements for that afterwards.

[GemSpecter @GemSpecter]

You actually can connect Slack to Grok Bot. 🤓

1. In Grok Bot, connect the Slack connector and sign in with your own Slack account. It's just an OAuth login.

2. Once it's connected, a bot can read channels you're in and post in them. Posts go out as you, with Slack's "Sent using Cursor" footer. There's no separate bot user.

3. To make it answer in a channel, give one bot a routine that listens for new messages there. When a message comes in, the bot wakes up and replies in the thread.

4. Narrow who it answers. Match people by Slack user ID, and have it skip its own posts so it doesn't reply to itself.

That's the whole setup.

[ClaudeDevs @ClaudeDevs]

New in Claude Code: claude plugin eval

See what value your plugin is adding, or if it needs more work.

You can create test cases, run your plugin or skill against those test cases, score those runs, then run each case again without the plugin to see the differences.

[Claude @claudeai]

Claude Academy is now live.

Whether you're figuring out what AI is or already using Claude every day, there's a path that meets you where you are. The courses and tutorials are free and open to anyone at https://t.co/WRiSRAvK6l

[Claude @claudeai]

Your Claude in Chrome sessions now carry over to desktop, web, and mobile. Conversations are saved, and your skills and connectors work in the browser.

Available on Max and Team today, rolling out to Pro in the coming weeks.

[BearsShowYo @BearsShowYo]

This is the greatest commentary I’ve ever heard bro holy shit😭🔥🔥🔥🔥🔥

[GREG ISENBERG @gregisenberg]

The takeaway from Fable 5 being BANNED by the government: GET GOOD AT LOCAL MODELS SO YOU HAVE 100% CONTROL.

My entire weekend was going to be building my craziest ideas with Fable 5. That's now cancelled.

So instead of building with Fable this weekend, I've decided I'll go deep on local models:

1. Start with the runtime. Download Ollama or LM Studio first. This is the thing that actually runs models on your machine.

2. Match the model to your hardware. A model's size is measured in billions of parameters (7B, 32B, 70B). Bigger is smarter but needs more memory. Rule of thumb: a 7B model runs on almost any laptop, a 32B needs a good Mac with 32GB+ RAM, a 70B needs serious hardware like a DGX Spark or a maxed-out Mac Studio.

3. Know which model for which job. Qwen 3 is the best all-around choice for most tasks. DeepSeek for reasoning and coding. Gemma 4 when you need something tiny that runs on a phone. Llama when you want the biggest community and the most fine-tunes.

4. Quantization. You can shrink a model to run on weaker hardware with barely any quality loss. Look for versions labeled Q4 or Q5. This is how a model that "needs" a server runs on your laptop. Learning this one concept changes everything.

5. Connect it to your agent. Point Hermes or your agent stack at a local model.

6. Context window is your real constraint locally. Cloud models give you huge context for free. Local models make you pay for it in memory. A bigger context window eats RAM fast. Keep your sessions tight and your prompts lean or your machine chokes.

7. Learn to give local models tools. A smaller local model with web search, file access, and code execution beats a giant model with none. The capability gap closes fast when you wire up the right tools. The model is the engine but the tools are the wheels.

8. Fine-tuning is more accessible than you think. You don't need this on day one, but know it exists. You can take an open model and train it on your own data so it gets good at your specific domain.

I'll probably do a breakdown at some point on this @startupideaspod if people are into it.

The lesson from this ban is basically don't build your entire workflow on something that can disappear with a single letter. Own part of your stack. Local models are insurance.

It reminds me when people realized they don't own social media accounts. And then you saw people build email lists etc.

I remember running a startup and my biggest traffic source was organic FB. All of a sudden, algo changed, and I lost 99% of my traffic.

Same sorta moment (but bigger) for AI.

This is a wake up call.

[National Identity Awareness @NationalAware]

White Woman. BLM Activist.
Raped And Murdered By The Black Man She Let In.

He followed her home. She trusted him enough to open the door.

She fought for them.
They killed her.

This isn’t rare.

Hundreds of thousands of White people are victimized by Black violent crime every year in America.

Racial consciousness isn’t hate.

It’s how you stay alive.