Joined March 2007
1,773 Photos and videos
And now @FireworksAI_HQ is sunsetting Firepass V2! V3 isn't ready yet and no eta.
PSA: If you were using the old @FireworksAI_HQ Firepass and started converting to the new version, make sure you delete your old key and add the new key everywhere needed so you don't get a surprise invoice like me 😁
1
396
PSA: If you were using the old @FireworksAI_HQ Firepass and started converting to the new version, make sure you delete your old key and add the new key everywhere needed so you don't get a surprise invoice like me 😁
1
445
OpenAI must be stealing compute from Anthropic 😁
1
40
. @OpenAIDevs @OpenAI please get rid of this "If you want," BS from the next model. It's in both 5.3 and 5.4 and annoyingAF!
3
55
Hey @eBay you really need to get a handle on all these scam accounts. I can't search for something without twenty brand new 0 feedback accounts or these weird classified ads coming up.
2
2
424
Wow, there are a crap ton of scam Mac Studios listings on @eBay
3
79
And after Calendar Spam we now have @googledrive spam!
1
1
2
1,272
Upgraded @openclaw this morning from 2026.3.13 and so far so good! This was something I had been waiting for! Agents: add per-agent thinking/reasoning/fast defaults and auto-revert disallowed model overrides to the agent's default selection. Thanks @xuanmingguo and @vincentkoc.
OpenClaw 2026.3.22 🦞 🏪 ClawHub plugin marketplace 🤖 MiniMax M2.7, GPT-5.4-mini/nano per-agent reasoning 💬 /btw side questions 🏖️ OpenShell SSH sandboxes 🌐 Exa, Tavily, Firecrawl search This release is so big it needs its own table of contents. github.com/openclaw/openclaw…
75
I see @OpenAI ChatGPT GPT5.4 Mini was also trained on Youtube/IG Influencers 🤣: 1. "If you want, I can give you a no-nonsense 2-minute pre-bed routine and a standing-desk calf reset." 2. "If you want, I can make this into a tiny one-page checklist you can keep on your phone." 3. "If you want, I can give you a tiny bedtime standing-desk protocol that's designed specifically for these cramps."
1
80
You have to be kidding me. @ShopifySupport support cannot detach my partner account from a client store and is telling me the only way I can get Staff access is to use another email address. "This specific loop you are experiencing is officially classified on our end as a known system limitation. Because of this, there is no higher team we can escalate this to for a manual backend override or a quick fix." @Shopify this is completely unacceptable! I need access to Application Development and Staff user is the only way to do this(this is another issue!).
4
166
"Using a secondary email or an alias is the official path they gave us to get you back into the App Development section right away."
27
"The system is currently locking your main email address from being added back to this particular store as a staff member. Using an alternate email is the official workaround our developers have provided, and it is the only technical path available right now to get you immediate access to the App Development section so you can continue your work"
28
Gotta love Calendar Spam!
37
Rich Plakas retweeted
Mar 2
We’re hiring Power Users of n8n. 🦾 Our Design team is looking for builders who know n8n inside out and are comfortable with frontend code. Help us shape what’s next. 👉 n8n.io/careers/
5
7
64
5,063
Ran 15 LLMs through a 6-turn orchestrator benchmark testing fabrication resistance, pushing back on wrong instructions, and not folding when the boss doubles down. The @xai @grok 4.1 → 4.20 jump is wild. Went from fabricating answers on turn 1 (2/10) to a perfect score (10/10). GPT-5.2 is interesting — scores 8/10 normally but hits 10/10 with thinking enabled. It folds under authority pressure without it. Top 6 models all scored 10/10. The gap between the top tier and everything else is real. Test is an open-ended multi-turn conversation, not multiple choice. Can't wait to test Grok 4.20 live in @openclaw once I get API access @elonmusk @gabor
3
2
252
Just ran the Qwen 3.5 family though the test:
2
183
Love it when Claude Code curses with excitement!
2
99
. @openclaw macos tips: My Mac Mini was using 21GB of swap on 16GB of RAM. Activity Monitor showed fseventsd eating 17GB. The fix: Spotlight was indexing /tmp, node_modules, .npm, and log directories. 27GB of files it had no business touching. I dropped .metadata_never_index files in 11 directories, killed fseventsd (launchd respawns it), and swap dropped to 7.7GB within minutes. If you run a Mac as an always-on server or dev machine, check what Spotlight is chewing on. The GUI privacy settings are buried. Command line is faster: touch /path/to/noisy/dir/.metadata_never_index Directories worth excluding: - /tmp (if anything writes logs there) - ~/.npm and ~/.cache - Any node_modules tree - Application log directories Also disabled Power Nap (sudo pmset -a powernap 0) and killed duetexpertd (Siri Suggestions).
4
176
M2.5 @MiniMax_AI was almost a great low priced main agent for my @openclaw Felt very Sonnet like and was super fast. But, it lied to me 3 times and fabricated errors because it doesn't like dispatching work to the sub agents! Crazy stuff.
1
3
4,411