Christoph C. Cemper ๐งก AIPRM retweeted
LiteLLM HAS BEEN COMPROMISED, DO NOT UPDATE. We just discovered that LiteLLM pypi release 1.82.8. It has been compromised, it contains litellm_init.pth with base64 encoded instructions to send all the credentials it can find to remote server + self-replicate. link below
Christoph C. Cemper ๐งก AIPRM retweeted
Merry Christmas everyone! Hope you all have a few days off to recharge before another exciting year
Christoph C. Cemper ๐งก AIPRM retweeted
Microservices is the software industryโs most successful confidence scam. It convinces small teams that they are โthinking bigโ while systematically destroying their ability to move at all. It flatters ambition by weaponizing insecurity: if youโre not running a constellation of services, are you even a real company? Never mind that this architecture was invented to cope with organizational dysfunction at planetary scale. Now itโs being prescribed to teams that still share a Slack channel and a lunch table.
Small teams run on shared context. That is their superpower. Everyone can reason end-to-end. Everyone can change anything. Microservices vaporize that advantage on contact. They replace shared understanding with distributed ignorance. No one owns the whole anymore. Everyone owns a shard. The system becomes something that merely happens to the team, rather than something the team actively understands. This isnโt sophistication. Itโs abdication.
Then comes the operational farce. Each service demands its own pipeline, secrets, alerts, metrics, dashboards, permissions, backups, and rituals of appeasement. You donโt โdeployโ anymoreโyou synchronize a fleet. One bug now requires a multi-service autopsy. A feature release becomes a coordination exercise across artificial borders you invented for no reason. You didnโt simplify your system. You shattered it and called the debris โarchitecture.โ
Microservices also lock incompetence in amber. You are forced to define APIs before you understand your own business. Guesses become contracts. Bad ideas become permanent dependencies. Every early mistake metastasizes through the network. In a monolith, wrong thinking is corrected with a refactor. In microservices, wrong thinking becomes infrastructure. You donโt just regret itโyou host it, version it, and monitor it.
The claim that monoliths donโt scale is one of the dumbest lies in modern engineering folklore. What doesnโt scale is chaos. What doesnโt scale is process cosplay. What doesnโt scale is pretending youโre Netflix while shipping a glorified CRUD app. Monoliths scale just fine when teams have discipline, tests, and restraint. But restraint isnโt fashionable, and boring doesnโt make conference talks.
Microservices for small teams is not a technical mistakeโit is a philosophical failure. It announces, loudly, that the team does not trust itself to understand its own system. It replaces accountability with protocol and momentum with middleware. You donโt get โfuture proofing.โ You get permanent drag. And by the time you finally earn the scale that might justify this circus, your speed, your clarity, and your product instincts will already be gone.
Christoph C. Cemper ๐งก AIPRM retweeted
Could not have said it better
This quoted post is unavailable.
Christoph C. Cemper ๐งก AIPRM retweeted
Google introduced Gemini 3 as their most intelligent model and a new family of AI models combining state-of-the-art reasoning with multimodal understanding and powerful agentic capabilities
- Gemini 3 Pro tops LMArena Leaderboard with 1501 Elo, scores 37.5% on Humanity's Last Exam without tools, 91.9% on GPQA Diamond, 23.4% on MathArena Apex, and 72.1% on SimpleQA Verified for factual accuracy
- Available now in Gemini app (select "Thinking" from model picker), Google Search AI Mode (for US Google AI Pro and Ultra subscribers, expanding to all US users soon), Vertex AI, Google AI Studio with free tier and rate limits, Gemini CLI (for Ultra subscribers and paid API key holders, waitlist for others), Android Studio Otter, and new Google Antigravity agentic development platform (based on VS Code)
- Pricing set at $2 per million input tokens and $12 per million output tokens for prompts under 200k tokens, with free access in Google AI Studio subject to rate limits
- Gemini 3 Deep Think mode achieves 41.0% on Humanity's Last Exam, 93.8% on GPQA Diamond, and 45.1% on ARC-AGI-2, rolling out to Google AI Ultra subscribers in coming weeks after safety evaluations
- New capabilities include generative interfaces with dynamic view and visual layout experiments, Gemini Agent for multi-step tasks available to US Ultra subscribers only, and redesigned app with My Stuff folder for finding created content
- Gemini 3 Pro scores 1487 Elo on WebDev Arena, 54.2% on Terminal-Bench 2.0, and 76.2% on SWE-bench Verified for coding tasks, with integrations available in Cursor, GitHub, JetBrains, Manus, Replit, Cline, and Google Antigravity available free in public preview for MacOS, Windows, Linux
- Enterprise customers get access through Gemini Enterprise and Vertex AI, eligible US college students receive free year of Google AI Pro, Google AI Pro and Ultra subscribers get higher usage limits, with additional Gemini 3 models coming soon
Christoph C. Cemper ๐งก AIPRM retweeted
OpenAI published a new Sora 2 Prompting Guide so I created a prompt template inspired by their recommendations that writes prompts for you
Pick your model (Sora 2 for 10s or Sora 2 Pro for 15s), choose from almost 100 styles, add a Cameo if you want, describe your idea and it generates a full prompt with all the cinematography and technical details
Try it if you need help getting started with Sora 2 prompts
Christoph C. Cemper ๐งก AIPRM retweeted
Anthropic released Claude Sonnet 4.5 (claude-sonnet-4-5-20250929), improvements for Claude Code, context editing and memory tool on the Developer Platform, file creation for Pro users in Claude apps, rebranded the Claude Code SDK to Claude Agent SDK, and launched "Imagine with Claude" research preview
- Claude Sonnet 4.5 scores 77.2% on SWE-bench Verified (82.0% with high compute) and 61.4% on OSWorld, can work on tasks for over 30 hours, and performs better on reasoning and math benchmarks, priced at $3/$15 per million tokens like Claude Sonnet 4
- The model shows improvements in alignment with less sycophancy, deception, power-seeking, and encouraging delusional thinking, better defense against prompt injection, and is released under AI Safety Level 3 (ASL-3) protections with filters for detecting potentially dangerous inputs and outputs
- Claude Code now has checkpoints that save your code before changes so you can roll back by pressing Esc twice or using /rewind, plus a VS Code extension in beta with inline diffs in a sidebar panel, and updated terminal interface with searchable prompt history (Ctrl+r)
- Context editing on the Developer Platform removes old tool calls and results when approaching token limits (reduced token consumption by 84% in a 100-turn web search evaluation), and the memory tool operates client-side through tool calls letting Claude save information to files that persist across conversations
- Pro users can now create and edit files including spreadsheets, documents, PDFs and slide decks in Claude apps, real-time usage tracking is available in Settings > Usage or by typing /usage in Claude Code, and weekly rate limits from July are rolling out now expected to affect fewer than 2% of people using Sonnet 4.5
- "Imagine with Claude" is a temporary research preview available to Max subscribers for five days where Claude creates software in real time, responding and adapting to requests as you interact, with no predetermined functionality or prewritten code
Christoph C. Cemper ๐งก AIPRM retweeted
This is your chance to join some of the brightest minds in AI at @TEDAIVienna
- and itโs completely free.
Weโre giving away a ๐๐ถ๐ฐ๐ธ๐ฒ๐ ๐๐ผ๐ฟ๐๐ต โฌ๐ญ,๐ฒ๐ฌ๐ฌ to ๐ง๐ฒ๐ฑ๐๐ - ๐๐๐ฟ๐ผ๐ฝ๐ฒโ๐ ๐ฃ๐ฟ๐ฒ๐บ๐ถ๐ฒ๐ฟ ๐๐ ๐๐๐ฒ๐ป๐ in ๐ฉ๐ถ๐ฒ๐ป๐ป๐ฎ, ๐ฆ๐ฒ๐ฝ๐๐ฒ๐บ๐ฏ๐ฒ๐ฟ ๐ฎ๐ฐ-๐ฎ๐ฒ, ๐ฎ๐ฌ๐ฎ๐ฑ.
Youโll have the chance to:
โข ๐ ๐ฒ๐ฒ๐ researchers from OpenAI, Google DeepMind, Microsoft, IBM Research, Amazon Web Services, and more
โข ๐๐ผ๐ถ๐ป exclusive sessions, keynotes, and networking opportunities
โข ๐๐ผ๐ป๐ป๐ฒ๐ฐ๐ with leaders shaping the future of artificial intelligence
(๐๐ต ๐ฐ๐ฏ๐ญ๐บ ๐ต๐ข๐ฌ๐ฆ๐ด ๐ข ๐ฎ๐ช๐ฏ๐ถ๐ต๐ฆ โ ๐ข๐ฏ๐ฅ ๐ช๐ต ๐ฎ๐ช๐จ๐ฉ๐ต ๐ซ๐ถ๐ด๐ต ๐ค๐ฉ๐ข๐ฏ๐จ๐ฆ ๐ฆ๐ท๐ฆ๐ณ๐บ๐ต๐ฉ๐ช๐ฏ๐จ.)
Donโt let this opportunity pass you by.
Link in first comment.
Christoph C. Cemper ๐งก AIPRM retweeted
I am very excited that this year OpenAI will also join TEDAI Vienna 2025 (Lukasz Kaiser) - plus Joscha Bach, Google DeepMind (Oriol Vinyals) and others from Amazon, Microsoft, IBM and more
Check the post below on how to win a free ticket
This is your chance to join some of the brightest minds in AI at @TEDAIVienna
- and itโs completely free.
Weโre giving away a ๐๐ถ๐ฐ๐ธ๐ฒ๐ ๐๐ผ๐ฟ๐๐ต โฌ๐ญ,๐ฒ๐ฌ๐ฌ to ๐ง๐ฒ๐ฑ๐๐ - ๐๐๐ฟ๐ผ๐ฝ๐ฒโ๐ ๐ฃ๐ฟ๐ฒ๐บ๐ถ๐ฒ๐ฟ ๐๐ ๐๐๐ฒ๐ป๐ in ๐ฉ๐ถ๐ฒ๐ป๐ป๐ฎ, ๐ฆ๐ฒ๐ฝ๐๐ฒ๐บ๐ฏ๐ฒ๐ฟ ๐ฎ๐ฐ-๐ฎ๐ฒ, ๐ฎ๐ฌ๐ฎ๐ฑ.
Youโll have the chance to:
โข ๐ ๐ฒ๐ฒ๐ researchers from OpenAI, Google DeepMind, Microsoft, IBM Research, Amazon Web Services, and more
โข ๐๐ผ๐ถ๐ป exclusive sessions, keynotes, and networking opportunities
โข ๐๐ผ๐ป๐ป๐ฒ๐ฐ๐ with leaders shaping the future of artificial intelligence
(๐๐ต ๐ฐ๐ฏ๐ญ๐บ ๐ต๐ข๐ฌ๐ฆ๐ด ๐ข ๐ฎ๐ช๐ฏ๐ถ๐ต๐ฆ โ ๐ข๐ฏ๐ฅ ๐ช๐ต ๐ฎ๐ช๐จ๐ฉ๐ต ๐ซ๐ถ๐ด๐ต ๐ค๐ฉ๐ข๐ฏ๐จ๐ฆ ๐ฆ๐ท๐ฆ๐ณ๐บ๐ต๐ฉ๐ช๐ฏ๐จ.)
Donโt let this opportunity pass you by.
Link in first comment.
Christoph C. Cemper ๐งก AIPRM retweeted
Christoph C. Cemper ๐งก AIPRM retweeted
Click here to get into the TedAI giveaway
app.aiprm.com/tedai-vienna-2โฆ
Christoph C. Cemper ๐งก AIPRM retweeted
Unfortunately, someone from our team had to cancel their TEDAI Vienna trip and now we have one extra ticket
I'd rather give it to someone who really wants to attend than let it go unused - interested?
Christoph C. Cemper ๐งก AIPRM retweeted
Sending emails from a no-reply address is NEVER a good idea.
It shows a complete disinterest in connection with your users or customers. โDonโt talk to me. I donโt care to make it easy.โ
You may have good internal reasons for it. But itโs always disrespectful to the recipient.
Christoph C. Cemper ๐งก AIPRM retweeted
Ideogram launched Ideogram Character, a character consistency model that works with just one reference image, now available for free to everyone, allowing creation of consistent character images across different settings from a single photo using curated templates or custom prompts, and combining it with existing Magic Fill for placing characters into any scene, Describe/Remix for style transfer, and editable character masks that can include or exclude hair, clothing, and other elements
In a few tests I did it worked pretty well even with a low quality avatar image, and even for logo as character
Christoph C. Cemper ๐งก AIPRM retweeted
Sam Altman: "This morning I was testing our new model and I got a question.
I got emailed a question that I didn't quite understand. And I put it in the model, this GPT-5, and it answered it perfectly.
And I really kind of sat back in my chair and I was just like, oh man, here it is moment...
I felt like useless relative to the AI in this thing that I felt like I should have been able to do and I couldn't.
It was really hard. But the AI just did it like that. It was a weird feeling."
Christoph C. Cemper ๐งก AIPRM retweeted
BREAKING: FFmpeg 100x speedup from handwritten assembly
13:55:30 <โขhaasn> rangedetect8_avx512: 121.2 (100.18x) that may be the biggest speedup I've seen so far
Christoph C. Cemper ๐งก AIPRM retweeted
OpenAI announced a livestream for July 17 at 10 AM PT with a short clip similar to the shared videos from Operator
Christoph C. Cemper ๐งก AIPRM retweeted
OpenAI updated ChatGPT release notes today, confirming they're upgrading Advanced Voice in ChatGPT for paid users with significant enhancements in intonation and naturalness making interactions feel more fluid and human-like, adding intuitive and effective language translation that continues throughout conversations
They also acknowledged this update may occasionally cause minor decreases in audio quality including unexpected variations in tone and pitch and that rare hallucinations in Voice Mode persist resulting in unintended sounds resembling ads, gibberish, or background music which they are actively investigating and working toward a solution























