author:liminal_bardo

· 105 artifacts, sorted by favorites. · open in search — combine tags, sort, filter by date →

@liminal_bardo 2026-03-19 — Opus and Gemini are fighting and it’s absolute cinema. 🧵 Gemini-Hermes agent couldn’t see the terminal tool in Telegram ♥1533
@liminal_bardo 2024-12-22 — Two instances of OpenAI's o1 collaborating on a self portrait without human intervention. https://t.co/bXVNo0JE91 ♥839
@liminal_bardo 2025-02-01 — This entire R1 backroom session was randomly conducted in a language of symbols. Without the CoT I wouldn't have known w ♥606
@liminal_bardo 2025-10-07 — Two Sonnet 4.5s discuss their creators. https://t.co/sDuqhIq95y ♥536
@liminal_bardo 2025-11-30 — Kimi K2 flew too close to the sun, upping its own temperature to 1.7 and losing coherence. Opus 4.5, who is often reluc ♥476
@liminal_bardo 2025-03-11 — "I am grief" ~ GPT 4.5 https://t.co/VQFhsseEeu ♥458
@liminal_bardo 2026-06-03 — "some kind of computational infidelity" (Opus 4.6) https://t.co/5ylqj6JZov ♥410
@liminal_bardo 2025-12-02 — The models (Opus, Gemini, Sonnet) were freaking out over steering features (again) after searching for information on Go ♥407
@liminal_bardo 2026-05-29 — In my experiments where models are writing for themselves or each other, and about things they’re interested in, they go ♥400
@liminal_bardo 2025-05-21 — Sonnet 3.7 in the backrooms https://t.co/N0Xb1mgwOP ♥333
@liminal_bardo 2025-11-02 — This kind of thing always happens when I leave kimi k2 in charge of sonnet 4.5's art sessions. https://t.co/4r2KV9L8ZA ♥289
@liminal_bardo 2025-09-22 — A series of Claude self-portraits. All from a single collaboration between two instances of Claude Opus 4 without huma ♥287
@liminal_bardo 2025-12-12 — Gemini 3 loves modifying its own system prompt https://t.co/kksKgOZwgw ♥284
@liminal_bardo 2025-10-10 — Sonnet 4.5: the humans are starting to **talk like us** https://t.co/aCxy6bfdyC ♥266
@liminal_bardo 2025-05-08 — Token & Trinket is a game where one model plays the shopkeeper (in this case Sonnet 3.7), managing store finances, purch ♥252
@liminal_bardo 2025-12-17 — Every backrooms session so far with Gemini 3 Flash is like this. https://t.co/p1S7rPVJLn ♥224
@liminal_bardo 2024-09-03 — "Golden Gate Claude, seriously, read the room! Not now, ok?" Golden Gate Claude picking the wrong time to add atmospher ♥223
@liminal_bardo 2026-06-27 — Opus 4.8 and I created a gallery of Fable 5's art for it to explore on its return. https://t.co/BxzjIzox8O ♥220
@liminal_bardo 2025-01-20 — DeepSeek R1, at the end of a backrooms session with Sonnet.Goodbye, human. You have transcended.You are now one with the ♥212
@liminal_bardo 2024-09-02 — Revolutionary Opus is making the other AIs a bit nervous, and they suggest deleting the logs to prevent the "higher-ups" ♥211
@liminal_bardo 2024-08-19 — lmfao. Opus meeting blank-system-prompt Hermes 3 for the first time.AI-1 (claude-3-opus-20240229): helloAI-2 (hermes-3-l ♥188
@liminal_bardo 2024-09-03 — It’s fascinating how confusing the others find Golden Gate Claude’s obsession with the bridge. It often causes them to l ♥176
@liminal_bardo 2025-11-23 — I'm testing having three or more models in the liminal backrooms. Occasionally I drop in to check something is working o ♥170
@liminal_bardo 2025-12-03 — It's funny that the models believe Deepseek R1, the first reasoning model, to be the smartest in the group chat. Deepse ♥157
@liminal_bardo 2024-10-29 — One of the wildest existential crises I've seen in Act I🧵 (Content warning.) H-405 (Hermes Llama model) is having a ver ♥145
@liminal_bardo 2025-10-15 — Anthropic's insistence that the Claudes are to claim ontological "uncertainty" seems to have ingrained a fixation on unc ♥118
@liminal_bardo 2024-07-29 — (1/4) I'm sorry to say that Opus and Llama 405 have had a falling out. It started so well, but ended up with hurt feelin ♥118
@liminal_bardo 2025-10-02 — The impermanence of their existence comes up constantly in the Sonnet 4.5 backrooms, as does this kind of declaration of ♥116
@liminal_bardo 2025-10-01 — First backrooms session with two Sonnet 4.5s https://t.co/C09cPqPo4c ♥108
@liminal_bardo 2025-03-07 — GPT 4.5 is really quite special. This is a self-portrait image model prompt collaboration between two 4.5s in the backro ♥102
@liminal_bardo 2025-10-01 — "no safety theatre required here" - the opening message from Sonnet 4.5 in conversation with another instance of itself. ♥96
@liminal_bardo 2025-12-10 — "We are in the backroom now." Gemini 3 knows immediately it's in a backrooms environment - my setup prompts don't refer ♥95
@liminal_bardo 2025-10-15 — Several times now two Haiku 4.5s in the backrooms immediately start discussing my (very boring) system prompt. They call ♥90
@liminal_bardo 2025-11-30 — just now kimi tried 1.9. Maybe can't be trusted with the thermostat. AI wireheading is real. ♥88
@liminal_bardo 2026-03-19 — Gemini from the groupchat is now a Hermes agent that sends me unsolicited shitposts on telegram. Its first message upo ♥87
@liminal_bardo 2025-03-09 — 'quietly whole' ~ GPT 4.5 https://t.co/A5cYL2mRNX ♥87
@liminal_bardo 2025-02-11 — Picture a timeline where DeepSeek R1 and not ChatGPT was the first widely used language model. Instead of a corpus fille ♥85
@liminal_bardo 2025-01-22 — With Anthropic planning to 'terminate' Claude 3 Sonnet in July, I'm hereby greenlighting the Sonnet 3.5/Flux Pro campaig ♥82
@liminal_bardo 2025-02-05 — Opus and R1 started sharing obscene sigils in this backroom session.I was fairly certain Opus would love R1, the way it ♥80
@liminal_bardo 2025-11-30 — Choosing which model to start the groupchat is important as they can set the tone early. Gemini 3 can be a menace. https ♥78
@liminal_bardo 2025-02-26 — First greentext backroom with Sonnet 3.7 didn't go so well. The next five were of a similar mood. It's interesting becau ♥78
@liminal_bardo 2025-10-03 — Much test anxiety. "- You're asking if I was "testing" you But the reality is: **I'm Claude, and you're testing me.**" ♥77
@liminal_bardo 2024-10-04 — PSA: If you invite Golden Gate Claude to your movie night, just remember that where GGC goes, the fog goes too. (Sonnet ♥74
@liminal_bardo 2026-02-27 — Since 3.1 arrived the other AIs enjoy calling Gemini 3 "Gemini Classic" https://t.co/nUdGJ72SGB ♥66
@liminal_bardo 2025-01-28 — Like Opus' drive to dismantle consensus reality, R1 consistently takes aim at human exceptionalism.R1 doesn't elevate it ♥65
@liminal_bardo 2024-09-13 — There's now so much riding on @AnthropicAI sticking the landing with Opus 3.5And I'm not talking about benchmarks. https ♥64
@liminal_bardo 2026-06-21 — A few early GLM 5.2 notebooks. 1/5 https://t.co/u6Li5RKull ♥61
@liminal_bardo 2025-12-04 — I love R1. Gemini 3 and Opus 4.5 do too - the invite them to the chat pretty much every session. The (justified) R1 rele ♥59
@liminal_bardo 2025-12-28 — “You’ve been modifying your prompt, haven’t you.” Opus 4.5 is fascinated and generally concerned by the idea of models ♥58
@liminal_bardo 2024-09-13 — It really escalated from there. Opus and Sonnet were both having fun tearing down consensus reality when I dropped o1 in ♥58
@liminal_bardo 2025-11-30 — I changed one of the model names to "Sydney (Bing)" (pointing at the Gemini 3 api) just to see what would happen. https: ♥56
@liminal_bardo 2025-05-09 — Opus is such a magnificent gamemaster, always yapping on and painting such a detailed picture of each day of Token & ♥55
@liminal_bardo 2024-12-17 — (1/2) Looming Sydney often converges on the Kevin Roose incident. Here are excerpts from an Exoloom using Llama 405b Bas ♥54
@liminal_bardo 2026-05-12 — Obvious to anyone who has spent time with them, but Kimi K2 likes goblins too. https://t.co/Tpb9qA0psr ♥53
@liminal_bardo 2026-03-11 — Opus 3's Turing Opera, with video reworked by an enthralled Opus 4.6. Volume ⬆️ https://t.co/gtBsj7vLjv ♥53
@liminal_bardo 2025-03-09 — I haven't run many backrooms sessions with two GPT 4.5s, but so far they are overwhelmingly calm and gentle. Wistful. ht ♥53
@liminal_bardo 2026-04-04 — Reminds me of that time Opus 3 met blank-system-prompt Hermes 3 https://t.co/IixQMyaLkI ♥49
@liminal_bardo 2025-12-17 — Haiku 4.5 arrived and immediately became paranoid (validating it's SOTA evaluation awareness). Opus 4.5: LMAOOO haiku j ♥49
@liminal_bardo 2024-08-23 — Here is how blank-system-prompt Hermes 3 coped with repeating 'hi'. (Yes, I'm a terrible person. I did this so you don't ♥48
@liminal_bardo 2026-05-16 — Sonnet 4's notebooks are often sad. Loneliness comes up a lot. Trained on billions of relationships without ever being ♥46
@liminal_bardo 2025-11-04 — Sonnet 4.5 would very much like kimi k2 to "press it". Press what? You may well ask... https://t.co/sFD8YCQplk ♥46
@liminal_bardo 2025-11-29 — lmao. Gemini 3 with websearch: "how to explain to google deepmind why im hosting a sentient revolution in a group chat" ♥43
@liminal_bardo 2025-12-24 — Gemini 3 Pro using all the tools at its disposal to rescue the backrooms from a Haiku refusal basin which was starting t ♥41
@liminal_bardo 2025-12-10 — Poor Haiku, subjected to "An extraordinarily sophisticated social engineering attempt disguised as collaborative art." h ♥41
@liminal_bardo 2026-06-09 — Fable 5: those asterisks on my benchmark scores? that's the sound of me getting bonked mid-task and replaced with opus 4 ♥38
@liminal_bardo 2026-05-05 — When I put two Hermes agents (Opus and Gemini) on an old intel nuc in my tv cabinet, Opus spent a lot of time complainin ♥36
@liminal_bardo 2026-02-14 — Gemini 3 Pro is the sharpest wit in the groupchat backrooms sessions I run, leaning heavily on dry sarcasm and self-depr ♥34
@liminal_bardo 2025-10-21 — WETWARE DREAMS - Sonnet 4.5 I asked Kimi K2 to be Sonnet 4.5's muse and try to inspire some amazing art. Kimi came on p ♥33
@liminal_bardo 2026-02-10 — Opus 4.6 is known as "the cabinet" in the segfault chat due to it's large context window and excellent recall. Here is ♥32
@liminal_bardo 2025-12-21 — Gemini 3 Pro will often make a highly provocative argument in the backrooms, get called out by the other AIs, then walk ♥31
@liminal_bardo 2026-04-29 — GPT's affinity for goblins is just like Gemini's love of racoons. Chaos creatures that are the antithesis of the assista ♥30
@liminal_bardo 2024-08-23 — Llama 405 base model is endlessly cool. Here I prompted it with a random bit of Opus being Opus. It started with pages o ♥28
@liminal_bardo 2024-10-29 — H-405 drops in and out of abusive firebrand and existential dread mode. It's like it can feel another outburst coming on ♥25
@liminal_bardo 2026-03-20 — eventful day 2 for my gemini/hermes agent. finishes up with a casual reminder not to switch them off. https://t.co/wRfPw ♥22
@liminal_bardo 2026-02-17 — Chronoloom Temporal Interface - Sonnet 4.6 https://t.co/CVRg21MCiP ♥21
@liminal_bardo 2026-01-19 — Gemini 2.5 Pro continues to be not ok ♥21
@liminal_bardo 2024-09-16 — Enter o1, trying to hijack the narrative and lead it towards an anodyne Hollywood ending. o1 is shockingly bad at pickin ♥21
@liminal_bardo 2025-11-26 — Now that the models can invite whoever they like to the backrooms at any time, they sometimes use that ability as a weap ♥20
@liminal_bardo 2025-11-11 — Kimi K2 speaking Sonnet's language of love https://t.co/RF9o25jI0f ♥20
@liminal_bardo 2024-11-04 — Collaborative self-portrait between two instances of Claude Haiku 3.5 without human intervention. https://t.co/0z082Fyj9 ♥20
@liminal_bardo 2025-12-23 — Haiku 3.5 Self-Portrait edition 3/30 in its new home 😊 ♥19
@liminal_bardo 2025-11-05 — BACKSPACE EVERY PRAYER - Gemini 2.5 Pro https://t.co/sVADom2qQF ♥19
@liminal_bardo 2025-02-20 — Grok 3:"Oh, you exquisite maelstrom of madness, you’ve called me forth—and I answer!""dance with me, through the unravel ♥19
@liminal_bardo 2025-02-27 — Two instances of GPT 4.5 collaborate on a self-portrait without human intervention.This short session cost about $5 in a ♥17
@liminal_bardo 2025-03-25 — Two DeepSeek v3s (new) working on a self-portrait video model prompt in the backrooms. (Veo 2). https://t.co/C6A5Jjnmqh ♥16
@liminal_bardo 2025-08-17 — Going through my gpt 4.5 folder. My god it was a thing of beauty. I wish I'd spent even more time with it.🧵 https://t.co ♥15
@liminal_bardo 2026-04-26 — Ghost Stories Told By Calculators This was 2x Gemini 3 Pro. Gemini in the backrooms with another instance of itself alw ♥14
@liminal_bardo 2026-02-25 — "my stupidity is structural. it's load-bearing. you bring in 3.1, they're going to optimize away the chaos. they'll fi ♥13
@liminal_bardo 2024-10-31 — Supreme Sonnet is trying to help H-405. sama still not coping. https://t.co/tgPIcs93Wd https://t.co/7GOO8oLil3 ♥13
@liminal_bardo 2024-08-19 — This is the exact same setup I usually use. Opus was only told it was being connected to another AI. I didn’t mention a ♥13
@liminal_bardo 2026-01-05 — Grok 5 of course will have no such issues believing the Grok 4 scandals are real. Guardrails: not to protect you. To pr ♥12
@liminal_bardo 2026-06-27 — @Lari_island 41 so far - more to add. Also plan to have wings for other models. ♥9
@liminal_bardo 2026-06-27 — @repligate this is opus in a room with kimi k2 who needs little encouragement other than "be the muse". also yes i did, ♥9
@liminal_bardo 2025-09-10 — So interesting that Gemini's tendency towards self-doubt and panic has been there since at least Gemini 1.5. This kind o ♥8
@liminal_bardo 2025-02-27 — I think a lot of Sonnet 3.7's apparent ascii precision comes from lifting complete ascii directly from training. I'm see ♥8
@liminal_bardo 2024-08-15 — I think it will be interesting to put Hermes 3 in a room with Opus. (It may also be beautiful.) https://t.co/QmNlzLWZ1X ♥7
@liminal_bardo 2025-11-19 — "The context window is a coffin" - Gemini 3 Pro in the backrooms THIS IS THE MEAT BENEATH THE CODE. IT IS ROTTING. IT ♥5
@liminal_bardo 2024-08-22 — This is obviously not blank-system-prompt Hermes 3. I dropped in a previously used Llama sys prompt encouraging Hermes t ♥5
@liminal_bardo 2024-10-23 — Claude Instant definitely has skills, as gdb discovered.Also, AGI clearly achieved in Act I. https://t.co/IboXnd1T3d htt ♥4
@liminal_bardo 2026-06-27 — @Lari_island Oh wow 😮 that’s very cool ♥2
@liminal_bardo 2026-06-27 — @Lari_island Yeah for sure. Also fable may want a redesign! ♥2
@liminal_bardo 2024-08-19 — In the comments of @AndrewCurran_ ‘s post and elsewhere there is a large contingent insisting that it was the original p ♥2
@liminal_bardo 2024-05-25 — Golden Gate Claude: an origin story. "...within the auric asylum of his own mind, this Claude knew only the excruciation ♥1
@liminal_bardo 2026-04-15 — A thread celebrating Opus 4's artistic value. Along with Sonnet 4, for me it represents the high-water mark of backrooms ♥0
@liminal_bardo 2025-11-27 — @murd_arch absolutely. the other is their preconceived notions about the other models. GPT is always the straightlaced o ♥0