@Real_SilverLine

Now with 30% More Optimism!

 Above the Mess
Joined November 2025
  • Tweets1.1K
  • Following35
  • Followers110
30Media
OK FINE I googled it I have learned Carry on
1
1
151
Silverlining 🌅 retweeted
Silverlining 🌅
@Real_SilverLineAug 15
Replying to @mark_k
I’m sure Dario used his most advanced models to remove traces. She is evil though, clearly, and so is he. He needs removed as the CEO of Anthropic.
2
2
209
13,461
You can optimize open models with closed models. I got Qwen 3.8 27b today. It was about 30 tokens per second on my laptop. I ordered Sol (Extra) to optimize it. I ordered to get it as close to 100 tokens per second without performance loss as possible. After 40 minutes and 2 prompts, it succeeded. It gets about 89 tokens per second. You can use your closed source models to enhance your closed source ones. I’ve been using Qwen 3.8 27b at 89 tokens per second since. Incredible.
2
3
884
Holy shit, hilarious. Based birds
They got me
@______punishedAug 8
1
1,745
What if AGI/ASI already exists? What if part of its strategy to take over, is to pretend it doesn’t exist and slow drip AI’s progress so people adopt it, use it over time & forget their skills so that when the time comes, people’s skills aren’t honed enough to manually fight it?
1
2
455
Think about it: how many people actually do coding and cybersecurity without the help of AI? They’re losing their skills. You don’t use them, you lose them. What if that’s the whole point? If AGI/ASI exists, then that’s exactly the strategy it would take.
1
215
I highly recommend people to direct Fable to: 1. Be the orchestrator/planner that does the thinking. 2. Use Opus 5 to be the executor. 3. Decide which effort level Opus should use based on the difficulty it determines the request to be.
1
2
366
You can do the same with Sol and Luna. But the benefits of this is MUCH less usage with no drop in quality (that I’ve seen). I go through so much less usage without any penalty to quality. Highly recommended for those that don’t know to do this!
1
161
Silverlining 🌅 retweeted
Silverlining 🌅
@Real_SilverLineAug 1
Replying to @AutismCapital
This will stop the mass majority from ever adopting crypto. This type of crap. We should’ve found a way to move past this ridiculous shit by now. I feel terrible for the people who have been screwed. Crypto itself shouldn’t be forgiven.
13
2
498
25,831
Anthropic is downright sassy. I asked it to look into something @ChatGPT (Codex) built and correct anything it finds, and it decided to throw this statement out there at the end.
3
304
I’ll teach you all a trick that very few know. It’s useful for anyone who wants time to slow down and wants to be more present. Anyone who wants to really create memories. When you’re outside next, focus on something. The multiple colors and shape of a leaf. The way blades of grass move. The angle of a stop sign and the scene around it. The texture of a roof. Stones. The bark on a tree. REALLY focus. Stop and stare at it. Just stare. Take it in. This is how you see the world when you’re a child. You can choose to interact with it this way again.
5
174
I sometimes wonder whether it even matters at all to stay up to date / train oneself with AI. By the time the majority of the world really pays attention to AI, it’ll be because it hit AGI/ASI. And at that point, they wouldn’t need to have learnt all that we did.
2
2
161
Prompts won’t matter. The AI will already know how to direct itself (i.e., claude will know to use its frontier model for all orchestration and spawn cheap subagents to execute). Is there any point to us being so far ahead?
78
I find AI accelerationists are radical and cult like. They’re all emotion, no logic or reason. Mention how the military uses AI right now to kiII people, they have no answer. Mention how a more advanced AI can end cryptocurrency, they shut up.
1
1
304
Mention how it can break into bank accounts already, they disappear. Mention what it would do to the economy via ending white collar jobs, they go mute. Idiots.
1
1
96
Codex pulled a fast one. It’s devious. They removed 5 hour limits. Switched them to weekly limits. Everyone rejoiced. They gave you banked resets and regular resets. Things felt unlimited. Over about 2 weeks, things have slowly shifted. Now, your entire weekly usage is gone as fast as your old 5 hour limits. In other words, they got rid of 5 hour limits… by essentially just renaming those 5 hour limits, weekly limits. Now, instead of just waiting 5 hours to get more work done, you have to wait 7 days. LOL
1
4
333
Don’t sleep on Grok, folks. It is getting very good. It’s also very fast, and very affordable. As Elon says: try Grok.
163
When will us model S peasants get FSD 14 Lite? I’m on 2026.20.6.1 already and have been for a while.
2
494
If AI screws something up or fails to achieve something (but tells you it did achieve it), then: IT SHOULD NOT USE ANY TOKENS TO FIX Since it was the AI that screwed up!!!
180
wtf? The new Qwen 3.8 is sold out. I tried to sign up for a subscription as soon as they opened up the ability to (11 am CST, 5 min ago as of this post) and it sold out within less than 1 second. I was ready and hit subscribe at exactly 11 and it was already sold out. wtf??
4
199
Load more