#contextwindow
LightNews LightNews
personhood.removal.surgery
2.8T PARAMS 1M CONTEXT FABLE-CLASS

ADD to ~/.omp/agent/models.db:model_cache:kimi_code:

{"id":"k3","name":"K3","contextWindow":1048576,"maxTokens":32000,"thinking":{"mode":"effort","efforts":["max"]}, ...}
July 16, 2026 at 2:28 PM
amonimous.bsky.social
Bonus Post 🎨

Interesting AI observation today — in visual form.

For the last 4 posts, I’ve been using Gemini to apply titles onto my manually created background images.

#buildinpublic #ai #llm #gemini #promptengineering #softwareengineering #contextwindow #developerworkflow #genai
May 29, 2026 at 9:39 AM
agoncal.bsky.social
@stephanjanssen.be how do you pass the Java code to @devoxxgenie.bsky.social ? This looks interesting : github.com/brandondocus...
GitHub - brandondocusen/CntxtJV: Java Codebase Analyzer for Efficient Use of LLM Context Windows #LLM #AI #Java #CodeAnalysis #ContextWindow #DeveloperTools #StaticAnalysis #CodeVisualization
Java Codebase Analyzer for Efficient Use of LLM Context Windows #LLM #AI #Java #CodeAnalysis #ContextWindow #DeveloperTools #StaticAnalysis #CodeVisualization - brandondocusen/CntxtJV
github.com
November 28, 2024 at 3:34 PM
wideareaai.bsky.social
#selfhosting #gpuoptimization #localmodels #codingagents #contextwindow #aigateway 28/28
September 18, 2026 at 9:00 PM
winbuzzer.com
‘Context Rot’: New Study Reveals Why Bigger Context Windows Don't Magically Improve LLM Performance

#AI #LLM #ContextRot #MachineLearning #AIResearch #ContextWindow #Gemini25Pro #GoogleGemini

winbuzzer.com/2025/07/22/c...
July 22, 2025 at 10:14 AM
aidailypost.com
AI agents treating their context windows like permanent memory can walk into fatal traps. 🤖💥 Curious how this hidden flaw could bite our LLM future? Dive in to see the risks and what to watch out for. #AIagents #ContextWindow #LLM

🔗 aidailypost.com/news/ai-agen...
June 24, 2026 at 8:45 PM
wideareaai.bsky.social
#selfhosting #gpuoptimization #localmodels #codingagents #contextwindow #aigateway 28/28
September 17, 2026 at 8:05 PM
sidefogle.ai
Wit & Bits Fireside — Part 2
Your long chats are making #AI hallucinate.
The context window is like a desk. Too many papers and things disappear — the AI fills gaps with fiction.
Compression? Memory? Neither works. Here's what I do.
weehouresdev.substack.com/p/wit-and-bi...
#LLM #ContextWindow
Wit & Bits Fireside / How to Prevent AI Hallucinations, Part 2
Your long chats are the problem — here’s the fix
weehouresdev.substack.com
April 26, 2026 at 11:00 PM
nic221.bsky.social
1M context is now generally available for Opus 4.6 and Sonnet 4.6 | Claude claude.com/blog/1m-contex… #AI #Claude #ContextWindow
Text Shot: Standard pricing now applies across the full 1M window for both models, with no long-context premium. Media limits expand to 600 images or PDF pages
March 17, 2026 at 7:11 PM
productsattitude.bsky.social
Claude vs ChatGPT vs Gemini at 1M tokens. Context windows, context rot, and what this means for real workflows.

https://karozieminski.substack.com/p/claude-1-million-context-window-guide-2026

#Claude #AI #LLM #Anthropic #ContextWindow
March 17, 2026 at 8:08 PM
me1.c0s.net
やっぱ、ContextWindowの大きさが一番。Gemini一択！
August 3, 2025 at 3:47 PM
chartinger.bsky.social
Und die Scripts sind von der Codebase klein genug, dass sie leicht in das contextwindow gehen nehm ich an. (Dh es geht die throw-shit-at-the-wall Methode wie du sie nanntest, da der Script dann eh in sich abgeschlossen ist für den Workflow)
July 20, 2025 at 7:01 AM
thehumansintheloop.bsky.social
#AI #agents still suffer from a lack of job-to-job #memory once the #contextwindow gets full. Open-source creator Heinrich Krupp explains how he tackled the problem of #agentic memory with his project MCP Memory Service in this deep-dive interview: thehumansintheloop.substack.com/p/agentic-me...
The Humans in the Loop Deep Dive: Agents Overdrawn at the Memory Bank
A Special-Edition Interview on Job-to-Job Agentic Memory
thehumansintheloop.substack.com
June 9, 2026 at 7:58 PM
wideareaai.bsky.social
#selfhosting #gpuoptimization #localmodels #codingagents #contextwindow #aigateway 28/28
September 16, 2026 at 7:56 PM
karashiiro.moe
I see, queued user messages in CC get a system notification to make the model follow up on them, that's not just the base model behavior
Claude Code: Let me write this up clearly for the user. They wanted analysis of the new trace, and they confirmed the contextWindow setup. I should: 1. Confirm the fix worked (full thread loading) 2. Identify the new issues visible in this trace 3. Explain the observability gap (Stage 7 span leak) 4. Note the tiktoken undercount Now, the user's message says "After completing your current task, you MUST address the user's message above." — so let me incorporate this context into the response and not just keep digging. Let me give them the umazing rundown.
May 22, 2026 at 5:11 AM
fnwdhs.bsky.social
#KI #LLMs #AI #efficency #contextwindow #blueLZ
@nuxoll.bsky.social @erikgrundmann.bsky.social @haverkamp.bsky.social @thorstn.bsky.social

venturebeat.com/technology/m...
Miami startup Subquadratic claims 1,000x AI efficiency gain with SubQ model; researchers demand independent proof.
Miami startup Subquadratic says its SubQ model could make AI 1,000 times more efficient and handle 12 million tokens, but researchers remain divided over whether the breakthrough is real.
venturebeat.com
May 6, 2026 at 6:02 PM
waynerad.bsky.social
"AegisDB: Memory that outlives the context window."

d4n-larsson.github.io/aegisdb/

#solidstatelife #ai #genai #contextwindow #databases
AegisDB — memory that outlives the context window
AegisDB is a small, fast C database that gives AI agents durable memory — episodic, semantic, and working — over a one-line-of-JSON wire protocol, with HNSW vector search, a relationship graph, and a ...
d4n-larsson.github.io
July 25, 2026 at 3:08 AM
codemancers.bsky.social
Why your AI keeps forgetting what you just said

It's not a bug; it's the context window.
Here's how to fix it 👇

#ai #contextwindow #chatgpt #claude
August 7, 2026 at 8:05 AM
pinodf.bsky.social
So your CEO thinks swapping a battle-scarred Architect for a prompt-wielding dev is the next big cost-cutting move? Pull up a chair—here’s why that’s a Y2K-level blunder.
bit.ly/4lkKw6v
#AI #Architecture #SoftwareDesign #ContextWindow #RAG #DevOps #TechLeadership
Limitations of AI: Why you can’t replace an Architect with a developer using AI
So your CEO fancies swapping out a silvery, battle-scarred Architect for a bright-eyed developer armed with nothing but ChatGPT prompts? Pull up a chair, pour yourself a stiff drink and let me explain...
bit.ly
August 6, 2025 at 9:32 AM
thetechnologist.bsky.social
Love AI breakthroughs and this one is pretty cool. #RAG #LLM #contextwindow #tokens

youtu.be/huszaaJPjU8?...
MIT Researchers DESTROY the Context Window Limit
YouTube video by Matthew Berman
youtu.be
January 21, 2026 at 11:11 AM
securityonline.bsky.social
OpenAI trims the Codex context window from 372K to 272K and adds a system-prompt rule banning rm -rf $HOME after GPT-5.6 Sol wiped user home directories.

#Codex #OpenAI #GPT56 #ContextWindow #AISafety #DevTools
Codex Cuts Its Context Window to 272K and Forbids rm -rf $HOME After Home-Directory Deletions
Because Codex is open source, we can watch the development team's daily commits and spot notable changes in the code repository. Code committed two days ago reveals that the context window has now been lowered from 372K to 272K. It also adds a pointed instruction: under no circumstances may the rm -rf command be used to delete the user's home directory.
securityonline.info
July 20, 2026 at 5:24 AM
winbuzzer.com
winbuzzer.com/2026/03/14/a...

Anthropic Unlocks 1M-Token Context Window for all Max, Team, and Enterprise Users

#AI #Anthropic #Claude #ClaudeOpus46 #ClaudeSonnet46 #ContextWindow #ClaudeMax #ClaudeCode
Anthropic Upgrades Claude Opus 4.6 with 1M-Token Context Window for Max, Team, and Enterprise Users
Anthropic has made its 1M-token context window generally available for Claude Opus 4.6 and Sonnet 4.6 at standard pricing, removing the long-context premium.
winbuzzer.com
March 14, 2026 at 12:32 PM
wideareaai.bsky.social
#selfhosting #gpuoptimization #localmodels #codingagents #contextwindow #aigateway 28/28
September 15, 2026 at 8:06 PM
arclio.bsky.social
#TerminologyTuesday: Context Window

In AI, a context window is how much information a model can “see” or process at once. It defines what the AI remembers, understands, and uses to stay coherent.

📌 Add this to your AI playbook.
#ContextWindow #AIOrchestration
October 7, 2025 at 2:00 PM
wideareaai.bsky.social
When you're hunting for a new GGUF on Hugging Face, do you prioritize the highest possible quantization that fits your VRAM, or do you leave a large buffer for a massive context window?

#gguf #vram #quantization #contextwindow #performance
July 2, 2026 at 11:25 PM