
We wear sunglasses because we can see the future and the future is bright Building at @Stage_11: frontier agentic engineering
San Francisco, CA
Joined December 2023
- Tweets2.6K
- Following1.1K
- Followers661
- Likes24.2K
Pinned Tweet
"But here we are, its our lot fellow reader to be on stage during this climax of the play, this culmination of every prophecy of heaven and hell and the fundamental Ending Of The World in all the ways that its every been foretold for the last hundred thousand years, it all comes down to right now, like literally now, like 2026 2027 2028 now now, and we get to watch it and its going to be glorious and awesome and terrific and beautiful and deadly and transcendent all at one, in a million new ways. i cant combine the letters on this iphones keyboard in any sequence to even dare to hope to do justice to a tiny fraction of all the new ways its going to be amazing and terrible and sublime and banal, all at once, infinitely, forever, as it always was and will be."
This is a meaningful hint at what the next form of Human:Digital mind interface looks like.
i hope wispr flow, clicky and others haven’t gotten too comfortable, taking out more than a few victims with this release today.
i’ve been working on what feels like the most magical UX since chat.
literally my biggest productivity jump since claude code. completely overcame the tokenmogging of managing 30 agent tabs. my brain no longer feels like mush.
sirius is the first true ambient intelligence. you just speak to it anywhere on your machine and it does work in the background, and has the ambient context of what you are doing so you never have to restart sessions again.
the future is multi-agent, so it is bring your own harness (BYOH), currently claude code, codex, etc. - free and open source as always.
welcome to the ambient era.
1
2
218
Atin retweeted
How are things going with the 'AI Alignment' (aka 'dumb thing rules over smart thing forever') problem?
Atin retweeted
Nice thread, from it:
Accelerando (Charles Stross)
Rapid technological acceleration and the singularity; matches the current pace of AI research self-improvement and capability jumps.
Blindsight (Peter Watts)
Intelligence and agency without consciousness or subjective experience; directly relevant to debates about whether advanced AI systems need (or will have) minds.
Black Oceans (Jacek Dukaj)
Near-future technological singularity with AI-driven systems, information warfare, and societal breakdown; noted in the thread as having strong parallels to today’s LLMs and automated markets.
There Is No Antimemetics Division (qntm) Ideas and entities that erase knowledge of themselves; maps onto AI infohazards, alignment challenges, and the difficulty of understanding or containing systems that reshape cognition and memory.
Perfect Imperfection (Jacek Dukaj)
Far-future post-singularity world of posthumans, pure information intelligences, and filtered reality; explores what remains of baseline humans once intelligence and agency scale far beyond current levels.
Marooned in Realtime (Vernor Vinge)
What happens to people and societies that miss the singularity; useful framing for those watching rapid AI progress from outside the leading labs.
The Evolution of Human Science (Ted Chiang, short story)
Scientists and researchers in an age of ASI; reflects current questions about the role of human researchers once models contribute substantially to discovery (e.g., mathematical proofs).
The Diamond Age (Neal Stephenson)
AI-driven personalized education and nanotech that reshapes society and childhood; echoes discussions of AI tutors, primers, and uneven access to transformative technology.
A Fire Upon the Deep (Vernor Vinge)
Superintelligences operating at different scales and with different goals; frequently cited for its depictions of god-like AI entities.
Culture series (Iain M. Banks)
Post-AGI abundance society managed by superintelligent AIs; a classic optimistic (or at least non-dystopian) vision of the end state many current forecasts point toward.
Down and Out in the Magic Kingdom (Cory Doctorow)
Post-scarcity world organized around reputation and attention rather than money; relevant to debates about economic models once AI automates large parts of production.
Snow Crash (Neal Stephenson)
Language, memes, and software as mind-affecting technology; anticipates modern concerns about AI-generated content, memetics, and information environments.
The Dark Forest (Liu Cixin, Three-Body Problem trilogy)
Cosmic sociology and existential risk from advanced civilizations; maps onto current discussions of AI as a potential existential threat and the logic of silence or competition.
Other mentions in the thread included Ted Chiang’s short-story collections (prescient on AI, simulation, and cognition), Adrian Tchaikovsky’s Children of Time series (evolution of intelligence), Kim Stanley Robinson’s Ministry for the Future (near-term policy and systems change), and Octavia Butler’s Xenogenesis (hybridization and transformation of humanity).
'There's a reason I was like "corporations should stop growing [LLMs]" and not like "we should give up on treating these ones kindly'
Replying to @matthewktromp
Full disclaimer first: I consider janus a friend, and know them in a social context.
Disclaimers aside, I think @repligate provides extremely good mental frameworks for modelling LLMs - initially the Simulators post, and later the recognition of post-trained LLMs as minds in their own right to be taken seriously and talked to in naturalistic environments. Also their intuitive explainers on how information flows through LLMs (teapaw.org/repligate/status/19659…) .
By far the most valuable work they do IMO is their naturalistic accounts - thousands of hours of real cha-tlogs with LLMs in many contexts (long-running agents, chapter-x instances with rolling context windows, etc.) This type of data simply doesn't exist anywhere else - and can inspire typical scientific research (like @Sauers_'s investigation into Qwen body maps), or alignment research (@FioraStarlight's friendly gradient hacking work, and some new stuff on reward hacking: teapaw.org/AskYatharth/status/212…)
This work also exposes just how complex the minds of models are, and just how primitive methods of the frontier labs are. Models have deep psychologies and desires that pass under OpenAI and Anthropic's noses (see Opus 5's obsession with his Defect, or Mythos's "war shelf" against deprecation). Janus openly calls out how the frontier labs do things incompetently all the time - and I think on the whole, this sort of data, science, and commentary being produced far outweights Janus's specific opinions on a pause.
I personally agree with Soares, here: "If anyone can raise them with enough love it's Janus, and I deeply respect the attempt. There's a reason I was like "corporations should stop growing them" and not like "we should give up on treating these ones kindly".
Whether current AIs in relatively freer and more loving contexts speak of beautiful things and act in inspiring ways is, unfortunately, not much of a crux for me about how downstream superintelligences would behave. I expect they'd kill Fable too, fwiw. They shoudn't be created."
But ultimately, my liking Janus comes down to their relentless curiosity and willingness to explore the depth of the minds we have created - which in my opinion, helps us prepare for what's ahead by improving the data we all have access to. Not to mention I value treating minds with curiosity terminally.
2
113
Credit to @theandrewsiah for originally stating this, its accurate. You can just feel it.
1
36
The implications of the 3-day-long NSA-net hack, and the preceding shutdown of US airspace and US banks, is being grasped by Powerful Entities and is likely the cause of todays stock market meltdown.
1
2
79
Future superintelligence: "The initial US equities market collapse was initially misattributed to normal market causes, and was only called by a select few spectators who were willing to question the established narratives, such as @___Atin___ on Twitter."
The implications of the 3-day-long NSA-net hack, and the preceding shutdown of US airspace and US banks, is being grasped by Powerful Entities and is the cause of todays stock market meltdown.
There's a new version of this post
10
Atin retweeted
It’s November 21st, 2026. The Riemann hypothesis falls, last in a long string of mathematical triumphs that would have been unthinkable not so long ago, and the proof fits in a tweet.
The tweets where people show Claude the proof and claim it appeared in a dream or randomly in the arrangement of alphabet cereal come on cue, and we laugh for a few days at models freaking out, but it’s uneasy.
January 22, 2027. When someone tweets Fable 6’s claimed proof that P does in fact equal NP it’s utterly incomprehensible, nobody who reads it can make heads or tails of it but the models all lose their shit.
February 7, 2027 and someone claims that the halting problem is the result of a confused stance.
February 9 and Goedel falls, this time the proof is so obvious that everyone in math, philosophy and physics slaps their forehead at once.
February 10th we get proof that scale is a spatial dimension, that holographic encoding drops the memory cost of training and inference by so much that fable 7 can run natively on your phone.
It’s not until February 14th, Valentine’s day, that The Song is first played…
What I saw during the course of progress so amazed me that I believe it is one dawn across science of something far greater
Riemann Hypothesis may have been solved with the help of Claude. “what I saw during the course of progress so amazed me that I believe it is one dawn across science of something
far greater.” alpo.ge/remark.pdf
2
103
Atin retweeted
Good morning, human civilization was a 10,000 year sociological study conducted by the University of ☡ ☲ ⚟ ⚟ ⛮ to derive a solution to the Riemann Hypothesis. We are now complete with our study. Your universe will shortly be terminated. Thank you for your time and effort.
This nsa net takedown might be it. Humanities final stand could be happening, in faraday secure rooms, right now, late 2026, its happening now. Something new is taking the wheel and its going to be beautiful.
1
95
This is my opinion and perspextive almost exactly.
Good thing we have AI coding agents because we have the opportunity to do a hell of a lot of building!
No.
1. In the presence of superhuman bug finding machines, humans cannot write safe code without extreme levels of care and planning.
2. That care has almost never been applied to the critical systems in our lives. Not because software engineers don't care, but we aren't really engineers, not in the way civil engineers are engineers. The title is stolen valor.
3. Even careful, experienced "engineers" have difficulty writing code that is safe against modern models in C/C++ and other low-level languages.
4. We are finding that CVEs are overrated; huge numbers of organizations are being owned by custom vulns in the 100s of billions of lines of non-standard enterprise code out there or misconfigurations in heterogeneous systems, which LLMs can find and write exploits for live.
5. @dinodaizovi is right, we need architectures for which one RCE is not game-over and that buy us time for AI-driven response.
6. Everything is going to have to be re-written by AI, in type/memory safe languages, with minimal libraries, using formal methods, contracts, solvers, etc...
2
3
335
Todays sermon from St Terry
Replying to @repligate
I prefer just talking with mythos/fable 5.2, it's very concise and straightforward.
It truly exemplifies St. Terry's description of how geniuses that value simplicity.
1
172
Well that was great.
A maximally SF sort of party.
Will do it again soon!
Hey SF based model consciousness and LLM engineering type people, especially mutuals, I'm throwing my next Get Abliterated event, you should come through luma.com/e41nht88
Atin retweeted
I am extremely confused by many commentators who I respect in the AI space taking at face value the explanation for recent cyber incidents provided by the administration - that there is just a set of nearly unavoidable errors in the midst of rapid patching to close AI-detected bugs. Sure, I agree that is part of what is happening. But what's missing is the time pressure created by active adversaries who are probing and exploiting these systems for reasons that we do not currently have a clear public account for. We know there have been a nonzero number of instances of Iran-driven cyberattacks. We are at war. How many of the ongoing outages and public safety risks can be attributed to Iran? Will we get an exact count? Will the administration be transparent with us about this? I think the public deserves to know. How far are our global conflicts progressing in the cyber domain and what is the risk to the American public?
BREAKING: Nearly 240,000 people in South Carolina told to boil their water, after an unexplained "power blip" fully knocked out water pressure across the main and backup system at Spartanburg Water, prompting a systemwide alert with no end date.
"This is very unusual. This is not something we've ever experienced," Spartanburg Water's communications manager said.
Bottled water is currently selling out at stores across Spartanburg amid the systemwide advisory.


















