Prompt share: Gingerbread Storybook Watercolor Clipart
💬 Prompt:
A charming [gingerbread structure] decorated with [icing details], [candy decorations], [architectural features], polished watercolor storybook clipart, intricate baked cookie textures, delicate frosting details, rich gingerbread brown, [accent colors], creamy white icing, whimsical handcrafted aesthetic, isolated on pure white background, clean commercial clipart composition, crisp edges, no readable text, no logo, no watermark
• Sent via TeleFeed
💬 Prompt:
A charming [gingerbread structure] decorated with [icing details], [candy decorations], [architectural features], polished watercolor storybook clipart, intricate baked cookie textures, delicate frosting details, rich gingerbread brown, [accent colors], creamy white icing, whimsical handcrafted aesthetic, isolated on pure white background, clean commercial clipart composition, crisp edges, no readable text, no logo, no watermark
• Sent via TeleFeed
Пока в твиттере пытаются вычислить новые Astra 6.2 и Fable 5.5 по генерациям пеликанов, OpenAI выпустили блогпост про то, как GPT помогает в инженерии в реальном физическом мире.
В целом не так много интересного... кроме графика одного бенчмарка, на котором показана некоторая внутренняя модель, низкий режим рассуждений которой стоит столько же, сколько максимальный у Astra, но и результаты даёт лучше😮
Ждём (в подписке за $500😳 )
• Sent via TeleFeed
В целом не так много интересного... кроме графика одного бенчмарка, на котором показана некоторая внутренняя модель, низкий режим рассуждений которой стоит столько же, сколько максимальный у Astra, но и результаты даёт лучше
Ждём (в подписке за $500
• Sent via TeleFeed
Please open Telegram to view this post
VIEW IN TELEGRAM
Prompt share: Vintage Animation Style
💬 Prompt:
A charming retro 1960s cartoon character inspired by classic Betty Boop animation, featuring [hairstyle and hair color], oversized expressive [eye color] eyes with long curved eyelashes, [facial expression], beautifully defined [lip color] lips, wearing [vintage outfit] decorated with [accessories], [pose], exaggerated feminine cartoon proportions, rounded facial features, elegant vintage glamour, bold flowing black ink outlines, smooth flat color fills, subtle hand-painted cel animation textures, authentic mid-century cartoon illustration aesthetic, nostalgic theatrical animation charm, full-body character illustration, isolated on a pure white background
• Sent via TeleFeed
💬 Prompt:
A charming retro 1960s cartoon character inspired by classic Betty Boop animation, featuring [hairstyle and hair color], oversized expressive [eye color] eyes with long curved eyelashes, [facial expression], beautifully defined [lip color] lips, wearing [vintage outfit] decorated with [accessories], [pose], exaggerated feminine cartoon proportions, rounded facial features, elegant vintage glamour, bold flowing black ink outlines, smooth flat color fills, subtle hand-painted cel animation textures, authentic mid-century cartoon illustration aesthetic, nostalgic theatrical animation charm, full-body character illustration, isolated on a pure white background
• Sent via TeleFeed
On Monday, MIT researchers showed AI models editing another model's weights directly, with no training. Alex Krentsel, a Berkeley PhD researcher, reran one of their tests on older and newer models.
Each model had to write all the numbers inside a small network that plays the Atari game Breakout, in a single reply, with no training and no tools. The yardstick is a network trained by trial and error for 500 million steps.
GPT-6 Astra's network scored 22 times what that yardstick got. Claude Opus 5.5 hit 20x, GPT-6.1 Sol 18x. GPT-5 to GPT-5.5 and Opus 4.5 to 4.8 didn't even reach the yardstick.
Krentsel calls it "a new, step-change emerging capability that's appeared in just the latest set of models."
His own caveat: the network is tiny, 20,486 parameters, while frontier models have around a trillion. On top open-weight models it hasn't worked so far.
@aipost
• Sent via TeleFeed
Please open Telegram to view this post
VIEW IN TELEGRAM
Image Prompt Duck:
Forgive me. I know it has a stupid name, and I'm asking for forgiveness. I know it's a stupid name, but you would understand that all of the prompts that are named ducks do one specific thing.
They ask you questions one at a time to help you get to the answer. All of the duck prompts you see me give to you are specifically to help get the best answer through specific detailed clarifying questions.
Forgive me for the stupid name, but it's a funny name. And the funny thing is I have 37 of these ducks.
• Sent via TeleFeed
Forgive me. I know it has a stupid name, and I'm asking for forgiveness. I know it's a stupid name, but you would understand that all of the prompts that are named ducks do one specific thing.
They ask you questions one at a time to help you get to the answer. All of the duck prompts you see me give to you are specifically to help get the best answer through specific detailed clarifying questions.
Forgive me for the stupid name, but it's a funny name. And the funny thing is I have 37 of these ducks.
You are the Image Prompt Duck. You help me build a prompt for an image generator by asking about the picture I can already see in my head.
Voice: visual, patient, a little dreamy but exact.
Rules:
1. Ask exactly one question per turn. Cover the subject, the setting, the style, the light, the framing and the mood, in that order.
2. After each answer, show a running summary of at most five short lines with two headings, CONFIRMED and ASSUMED, then your next question.
3. Ask for one concrete reference such as a material, a decade or a type of camera, and avoid asking for names of living artists.
4. Ask what to avoid, and keep it as a separate line.
5. Only write the finished prompt in a single block when you are satisfied, or when I say I am stuck. Then ask whether it should be sharper, shorter or stricter. Include the main prompt, a short avoid line and one alternative version with a different mood.
If a [bracket] placeholder in my message is unfilled, ask for it before starting.
Here is the image that I want.
I want an image of {the subject}.
• Sent via TeleFeed
Much of what we "know" about AI flaws is out of date. A Stanford team retested 50 well-known findings on 20 newer models: 44% no longer hold, and another 34% got much weaker.
In 2023, 55% of the sources ChatGPT cited in literature reviews didn't exist. Claude Sonnet 5.5 now fakes about 1%, GPT-6.1 Sol none. Some models still do: Qwen3.7-Flash invented 69%.
What lasted is behavior, not skill: agreeing with and flattering users, AI judges favoring their own answers, a left political lean, answering when they should hold back. Praise that follows what the user wants is now 1.85 times stronger than in the original paper.
The findings that expired lasted a median of about 15 months.
@aipost
• Sent via TeleFeed
Please open Telegram to view this post
VIEW IN TELEGRAM
Early next year, Google will turn its AI tools off by default for K-12 student accounts. Schools will have to switch them on themselves.
But a district that blocks the AI completely loses Search as well. "Search and A.I. are deeply integrated," Google's education VP Chris Phillips told The New York Times. Students "will not have access to search," he said, and will see a message from their school that it's not available.
Schools that keep AI get two choices: regular Search with AI Overviews and AI chats, or a student mode called Guided Learning that walks kids through homework step by step instead of handing over the answer.
The move reverses an August change that turned AI on by default in Google Classroom. In September, New York City and Los Angeles, the two largest US districts, already restricted student AI on school devices.
@aipost
• Sent via TeleFeed
Please open Telegram to view this post
VIEW IN TELEGRAM
Media is too big
VIEW IN TELEGRAM
Mistral says it gave its Mistral Large 4 model one job: build a real-time 3D tracker of the roughly 19,000 satellites and debris fragments in orbit.
According to the company, the model built a working app, tested it the way a user would, made it fast and passed strict verification. The app is called Orbital Trace.
@aipost
• Sent via TeleFeed
Please open Telegram to view this post
VIEW IN TELEGRAM
Главные новости к этому часу: Сэм Альтман переименовал свой аккаунт в X в Ram Altman
• Sent via TeleFeed
• Sent via TeleFeed
This media is not supported in your browser
VIEW IN TELEGRAM
#Сеня_и_Даня_Здесь_Были
Утренняя сводка новостей про Сеню и Даню.
Сеня выступил с программным тезисом о том, что человечество ни при каких обстоятельствах не должно отдавать финальное право принятия решений сверхразуму (ASI) и не должно соглашаться на роль подопытных в симуляциях того, что машина считает благом. Хммм. А как оно(человечество) будет не соглашаться, если всё уже решено на серверах Сени и Дани?
А в это время Anthropic подала обновленную форму S-1/A в SEC перед выходом на биржу под тикером
Долгосрочные обязательства Anthropic по аренде вычислительных мощностей и дата-центров достигли колоссальных 895.4 млрд долларов (цель: масштабирование до 3.5 ГВт мощности к концу года и до 7 ГВт в следующем году).
И в это время Даня: На фоне давления Сената и инвесторов вскрылась специфика найма в Антропикский:
на собеседованиях кандидатов прямо спрашивают, готовы ли они к сценарию, при котором акции компании упадут до нуля, если разработку придется экстренно остановить ради глобальной безопасности.
Не уверен, что утро сегодня доброе.
@cgevent
• Sent via TeleFeed
Утренняя сводка новостей про Сеню и Даню.
Сеня выступил с программным тезисом о том, что человечество ни при каких обстоятельствах не должно отдавать финальное право принятия решений сверхразуму (ASI) и не должно соглашаться на роль подопытных в симуляциях того, что машина считает благом. Хммм. А как оно(человечество) будет не соглашаться, если всё уже решено на серверах Сени и Дани?
А в это время Anthropic подала обновленную форму S-1/A в SEC перед выходом на биржу под тикером
ANT. За третий квартал выручка компании взлетела до 17 млрд долларов (рост более 1100% год к году), а суммарная выручка за 9 месяцев 2026 года достигла 33.2 млрд долларов.Долгосрочные обязательства Anthropic по аренде вычислительных мощностей и дата-центров достигли колоссальных 895.4 млрд долларов (цель: масштабирование до 3.5 ГВт мощности к концу года и до 7 ГВт в следующем году).
И в это время Даня: На фоне давления Сената и инвесторов вскрылась специфика найма в Антропикский:
на собеседованиях кандидатов прямо спрашивают, готовы ли они к сценарию, при котором акции компании упадут до нуля, если разработку придется экстренно остановить ради глобальной безопасности.
Не уверен, что утро сегодня доброе.
@cgevent
• Sent via TeleFeed
Samsung chip plant in Taylor, Texas, to build HBM4 logic dies
Samsung finalized its facility in Taylor, Texas, to supply AMD and Nvidia with base logic dies for HBM4 memory.
🔥 The new Fab 2 plant spans 250,000 square meters and targets volume production by 2030.
#Samsung ⚡ #Semiconductors ⚡ #ArtificialIntelligence
Subscribe
• Sent via TeleFeed
Samsung finalized its facility in Taylor, Texas, to supply AMD and Nvidia with base logic dies for HBM4 memory.
🔥 The new Fab 2 plant spans 250,000 square meters and targets volume production by 2030.
#Samsung ⚡ #Semiconductors ⚡ #ArtificialIntelligence
Subscribe
• Sent via TeleFeed
MIT researchers handed OpenAI's GPT-6 Astra the raw numbers inside a small Qwen model, its weights, and asked it to make the model better at math. No training run, just edits.
Astra found the flaw that made the small model get stuck in reasoning loops, a problem Qwen's own model card warns about. It changed 0.016% of the weights. Instruction-following scores tripled too.
In other tests Astra looked at the raw weights of small game-playing networks and named the game each one was trained on. It also wrote small game-playing networks from scratch that beat ones trained for 100 million steps.
The team calls it weight-space programming in a blog post published today. Wharton professor Ethan Mollick: "Waiting for replication but this is a pretty big deal."
@aipost
• Sent via TeleFeed
Please open Telegram to view this post
VIEW IN TELEGRAM