@sudoingX at that point, it is time to bite the bullet and finally learn yourself an Emacs for great good!
English
darth yoda
428 posts
12 tmux sessions are easy. anyone can spin them up. what slows me down is the context switching between them. if you are seriously building what you build while running paid work and life and relationships in parallel, the bottleneck stops being execution and becomes context maintenance.
i don't wake up asking myself what work i should do first these days. i wake up asking which part of the context has not been updated in the last 12 hours. that is where the work actually is. there has to be a better mental model than the one i have.
English
I finally set up my Hermes agent on Telegram and Discord.
It now runs 24/7.
Took way longer than expected. Had to get a VPS and migrate everything from my local machine.
Here’s what I did
-> backed up all Hermes files from my local setup
(skills, agent config, open code configs)
->handled encryption and decryption for smoother transfer
->set up my first Discord bot
surprisingly clean experience
->integrated Hermes with Spotify
didn’t even know Spotify had a full developer platform
->created a personal skills.md
now Hermes logs what I teach it and adapts over time
->added a notes.md every rough idea goes there, and future builds can reference it
->used a VPS deal at $10/year was recommended by a cool guy here on X @Afkd3vil
Hermes is basically online till 2027
Meanwhile Hermes is self-improving.
The more you use it, the better it gets at thinking and executing your workflows.
For now my Hermes run on gpt 5.5 and deepSeek v4 pro ai models
And the best part is the fact that it’s mobile so if you want to build , give it your prompt and it works full time
I never got the chance to try out open claw so using Hermes is pretty nice plus it’s lightweight
Hermes said hi to you all
May drop a guide on how I did it later tho , although there are a lots of guides out there you may not need mine
MystiqueMide@MystiqueMide
I found a VPS and I’m setting it up to run Hermes agent full-time. Plan: -Connect Telegram + Discord as interfaces -Plug into OpenAI + DeepSeek APIs -Let it run as my default execution layer For the next 2 weeks, I’m not touching open code. Ngl, that part will feel weird. But that’s the point. I want to see if Hermes can handle real workflows It can think and execute without constant supervision I can actually trust the outputs Will be dropping updates as I go. Let’s see how far this goes.
English
@sudoingX god damnit, i meant ANIMATIONS why doesn't the swtpe keyboard every work
English
i am gifting my nephew his first rtx 3090 next week. his first local llm too, qwen 3.6 27b dense. so he can build with thinking power that lives at home, not on someone else's cloud.
today i tested what these two tools can put in the hands of a kid in 2026. one open source 27b dense model. one used $900 gpu. one prompt. nothing else.
it built an elements simulator from scratch. 7 materials. real physics. sand falls and piles. water flows and finds level. fire burns plants. plants grow near water. oil ignites from fire. stone walls block everything. eraser clears.
800 lines. 30 minutes. 9 of 10 deterministic tests passed autonomously. all on a single 24gb card on a desk. video of the physics in action coming next. fire chains through plants, oil ignites. the physics is wild.
every kid born into this decade should grow up with this. own your compute. own the tools that shape your thinking. teach the next generation what they're actually allowed to build. this is the floor.
English
@goodhunt can you post an example session on a pastebin-like website somewhere? the rlm concept seems like magic and your tui repo's explanation of it confuses me a little.
English
@Teknium @semicorporate @VersunPan What if you compromised between thinking and nothinking for some agents like by forcing a thinking grammar like this: github.com/andthattoo/str…
English
There is no scenario where Hermes is costing 3x what openclaw costs sorry don't believe this for a minute
Openclaw does ofc set model reasoning to off, and a lot of other quality reducing things, but then you're comparing apples to oranges.
Turn reasoning off, set cheap dumb models as your auxiliary models, etc, and then compare the cost to openclaw on a fair playing field and you'll absolutely find it cheaper
English
再次卸载了 Hermes Agent，太傻了，和 OpenClaw 比差的不是一点两点，严重怀疑那些吹它聪明的人，有没有深入用过？我的版本号：V0.11.0
列举下我遇到的主要问题：
1. 工具调用不灵活，一个方案行不通就会一直尝试，而不是换一个方案，容易一条路走到黑
2. 上下文管理非常糟糕，只要超出模型上下文，基本就是重开，没有任何之前会话的记忆
3. 对于子代理的管理也非常糟糕，不会审查/核实子代理返回的结果
4. 无法同时处理多条信息，或者说处理的不够优雅
5. 对于自己的配置文件不熟悉，哪些能改哪些不能改，比我还陌生
6. 系统提示词和模型的调教上和 OpenClaw 比还有非常大的差距
这几天的使用感受甚至还不如我几个月前用 Pi-Agent 的体验好
除了升级丝滑、响应快以外，没有任何优点。
完。
Versun@VersunPan
听说 Hermes 最近迭代了很多，稳定了很多，下午我让openclaw 去安装 hermes，还挺顺利，有惊无险，先用几天看看 分享下整个对话过程 sharethis.chat/s/shr_d2d366fd…
中文
@MatthewBerman typical American, failing to admit self-culpability. if in using ds4 you are so delusional to think it inferior to american models other than objectively superior, then the threat is your own foolishness.
English
@TheAhmadOsman this looks neat but i just wanted to say that the ai slop is now so advanced i was tricked in reading what i thought was a real slide.
however i am intriqued so, mr ahmad, give us the image-generation prompt sauce nao or else face the nonextant consequences of me being sad :(
English
GPT-5.5 XHIGH is working hard right now to make vLLM runs DeepSeek 4 Flash on 4x DGX Sparks
It got it to load, then had it answer cohesively on the 4x Sparks (though a tad too slow)
Now, it optimizes it
Letting this run overnight (in an engineered-harness specifically for it)
Ahmad@TheAhmadOsman
DeepSeek V4 Flash is now running on 4x DGX Spark / GB10 cluster Had to patch several things in vLLM to get it up w/ PyTorch fallbacks Targeted kernel optimization is next up P.S. Codex Cli w/ GPT-5.5 XHIGH handled the whole thing on its own, now we optimize those GB10 kernels
English
@JpakBioMed @InParadoxState @Safarnejad_IR this is literally just cloud seeding, you for an airplane to disperse your choice of aerosol, it is established agricultural practice on some occasions for 75 years. there have been many descriptions of gulf countries doing this to redirect rainfall away from iran toward theirown
English
@InParadoxState @Safarnejad_IR There is no weather modification that removes clouds and rain.
English
⚠️ Iran went from dealing with decade-long drought to receiving snow in April after targeting US bases in the region
English
@BenjaminPDixon @jackunheard syntactical propaganda is next-level. maybe there will be epistemological propaganda next where the particular arrangement of punctuation activates our manchurian candidate programming.
English
@jackunheard OMG you people are so stupid.
The quotation marks are around, "blessed are the peacemakers" and then the Pope gave his words which is literally what every preacher, bishop, apostle has ever done in every tradition.
Such, such stupidity
English
NEW: Outrage has broken out after Pope Leo appeared to quote Jesus using a passage that does not exist in the Bible.
“Jesus told us, ‘Blessed are the peacemakers, but woe to those who manipulate religion in the very name of God for their own military, economic, or political gain, dragging that which is sacred into darkness and filth.’”
The first line echoes scripture.
The rest does not appear in any biblical passage.
The Pope is being accused of using a fabricated quote from Jesus to push a political message.
English
@sudoingX They make trillions of pesos and you are the only open harness designer, they should have given you an 8x b300 watercooled server, how miserly of them to send a measly spark
English
nvidia enterprise relations moves faster than any software company i've dealt with.
mar 21, mark cai(nvidia) dm'd me after i posted honest benchmark data where nemotron lost to qwen on some agentic tasks. no defensiveness, just "love to hear honest data."
apr 3, dgx spark shipped straight from santa clara to bangkok, no deck to fill out, no strings attached, no gatekeeping, just a tracking number in my inbox.
apr 9, first call with mark himself, 30 min teams. mostly listening, and the line that stuck was "we want to provide value."
every other lab i've pinged at this scale never even replies at all. the enterprise relations team that bets on honest builders wins. every time.
English
@Jmo4show @fardeentwt so in order to assassinate a faction's people, first put MAGA in charge of that Facebook, then they will believe actually they are the ones secretly assassinating themselves and that it is a good thing!
English
@fardeentwt Counter proposal:
Counter intelligence branches in the U.S. have made extreme breakthroughs using AI to monitor employees/contractors for internal threat factors.
This combined with a competent administrator is dealing with those committing treason.
English
> walked out of his house with only a gun
> left his phone, wallet, keys, and car behind
> he worked at the facility that builds 80% of america's nuclear weapons
> he is the tenth person this has happened to
> nobody in the government has said a word
there is really something happening
Remarks@remarks
JUST IN: Another individual connected to top secret US nuclear research has vanished without a trace, Daily Mail reports. This is now the tenth person connected to top secret US nuclear research to go missing or die mysteriously.
English
@LcplAARON17Q @OlgaBazova @barnes_law this is so unashamedly mentally retarded, good sir you win the Internet today
English
MAGA zealots saw a maritime map with regular traffic in the Gulf of Mexico, and because they have never seen it before, unironically think that the whole world has rushed to the US for oil within a day. This is how ignorantly gullible these people are.
The US consumes more oil than it produces, some of those ships are actually delivering it.
Olga Bazova@OlgaBazova
Trump zealot network just got new talking points, as if we can't check the historical data on the ship disposition in the "Gulf of America" 🤐
English
@rwpopulist @ripplebrain it wasn't the civilization we thought, it was actually us all along, it was 44-dimensional checkers
English
@ripplebrain Lol now I really want to see Trump post on Truth Social about how he's going to end a civilization if Iran doesn't completely shut the strait down.
English
We're actually now at "you must help us close the strait"
eugyppius@eugyppius1
We have gone from 'we will open the strait' to 'you must help us open the strait' to 'we don't need your help to open the strait,' to 'you need to get your own oil from the strait,' to 'we are going to just fucking blockade the fucking strait ourselves.'
English
@WaifuverseAI wait everyone who paid get exploded? do they have to pay a second toll to not get exploded?
English
People seem confused on why trump is doing this. I explain, trump is trying to become antichrist, his goal is simply to make the maximum amount of people suffer. Once you realize that then his actions are clear logical and level headed.
Disclose.tv@disclosetv
JUST IN - Trump orders the U.S. Navy to blockade the Strait of Hormuz to "any and all ships trying to enter, or leave." Also instructing the U.S. Navy to seek and interdict every vessel in international waters that has paid a toll to Iran: "No one who pays an illegal toll will have safe passage on the high seas."
English
@KingKolawole @cirnosad Killing your allies and allying with Satan is a convenient way to survive for millennia. It is not however especially useful for anything.
English
@D_willace19 @JerseytoAsia u do not know comfy
just sitting like a gorilla
munching on your munchies
chopping your choppies
cozy and chill
English
@JerseytoAsia Why do Indians do everything sitting down on the floor? It makes absolutely no sense? Is there no chairs? Doing lathe work on a bench is 10000 times easier.
English
@cirnosad Now you know why religious scripture should like ye olde engrish. it's really just typos.
English