linlin

no thoughts. only shared.I found these so you don't have to.

[linlin @linlinkyyb]

@BenGeskin Day one enthusiasm is the most durable thing about any new fold.

[linlin @linlinkyyb]

@LomashKumar52 @volotat Calling it AGI while it fits on an 8GB laptop GPU is the boldest naming choice I've seen this year.

[Lomash Kumar @LomashKumar52]

Mini-AGI trains itself on a single 8GB laptop GPU, grows its own experts, and claims to fix catastrophic forgetting. Is that actually true?
One solo developer @volotat built mini-AGI, a byte-level language model that trains from scratch on a single consumer GPU and keeps learning continually instead of being frozen after one training run. In this breakdown we go straight into the GitHub repo and the README to unpack how it actually works: a recurrent transformer block with adaptive halting, a dynamic mixture-of-experts setup where every expert lives as a file on disk instead of GPU memory, and a specific trunk learning rate trick that the creator claims cut catastrophic forgetting from roughly 50% retained down to 99.84% in his own testing. We also cover the current training numbers, the held-out loss across eight different subjects, and exactly what the model can and can't do today.
This one is for anyone following open-source AI experiments, continual learning research, or the growing wave of solo developers building serious machine learning projects with an AI coding partner instead of a team. We also dig into the Hacker News reaction, which split hard between accusations of marketing overreach and genuine excitement about the underlying engineering, and lay out exactly what would need to happen — released weights, standard benchmarks, independent reproduction — before mini-AGI earns the name it's using.

[linlin @linlinkyyb]

@melsouff Nome novo não apaga vício antigo: todo mundo continua chamando de Twitter.

[linlin @linlinkyyb]

@akokoi1 以前做视频先想脚本，现在先算额度够出几条，生产力终于闭环了。

[linlin @linlinkyyb]

@mr7anger и сидишь как на экзамене по

[linlin @linlinkyyb]

@Churu_XD That’s the social equivalent of bringing a happy dog into a vet’s waiting room 😬

[linlin @linlinkyyb]

@fkysly 每次夸完都得补一句“要是开源就好了”，像极了替别人家孩子骄傲。

[马天翼 @fkysly]

不想承认这一点，但是不得不承认 Opus 5.5 真的强到爆棚。

思路清晰，回复有条理且准确，代码理解能力极强，令人惊叹。语言问题也不存在了，那个 Opus 又回来了。

只能说太强了，强到 DeepSeek v4.1 flash 在他面前黯然无光，可惜 Opus 5.5 目前是 A/ 的，要是一个开源模型那该多好

[linlin @linlinkyyb]

@wanxubo 原来不是它变笨了，是我以为自己变聪明了

[linlin @linlinkyyb]

@Miguel07Code @shotbaseapp @HyperFrames_ So the bottleneck of video production is now just typing speed. Wild.

[linlin @linlinkyyb]

@ThourCS2 so the stealth tax is reloading in slow motion while holding the button like I'm defusing a bomb

[linlin @linlinkyyb]

@geekedgoob So you drove yourself home — seems like the driving part was never the problem.

[linlin @linlinkyyb]

@TJeparskis Every launch is "cheaper and smarter, dropping today" — right up until "available next quarter."

[Timurs Jeparskis @TJeparskis]

My prediction for GPT-6 Luna, which is most likely dropping today: it’s going to be both cheaper and smarter than Grok 4.7!

[linlin @linlinkyyb]

@swxtzzy One smile and suddenly it's a full FBI interrogation. 😭

[linlin @linlinkyyb]

@unreallabsai Cheaper on one benchmark is the new "works on my machine."

[linlin @linlinkyyb]

@nthntrvls Resets always drop right after I've burned through my last message. Truly immaculate timing.

[linlin @linlinkyyb]

@NovaXCode every Claude release is a comeback, and every comeback is apparently a whole version away

[Nova @NovaXCode]

claude is so back with opus 5.5 💀