Jerry Blount (@jerryblount) header
Jerry Blount (@jerryblount) profile picture

"We can't solve problems by using the same kind of thinking we used when we created them." - Albert Einstein

Joined October 2008
Photo by Jerry Blount (@jerryblount)Photo by Jerry Blount (@jerryblount)Photo by Jerry Blount (@jerryblount)Photo by Jerry Blount (@jerryblount)Photo by Jerry Blount (@jerryblount)Photo by Jerry Blount (@jerryblount)Photo by Jerry Blount (@jerryblount)Photo by Jerry Blount (@jerryblount)Photo by Jerry Blount (@jerryblount)Photo by Jerry Blount (@jerryblount)
I just realized that neither Codex desktop nor Codex CLI support the 1M context window, even with a Pro account. Perhaps I missed it on the marketing pages, my initial reaction, "I've been scammed", 5.6 Sol thinks it's a bit misleading @OpenAI ๐Ÿคท
4
1
140
@thsottiaux nitter.cf/thsottiaux/status/2089โ€ฆ ...The server will ignore this. The UI will in the desktop app and the /status in the CLI will show the 1M (actually a little less because of reserved token), but the server will overwrite it, without telling the frontend.
Here is how to enable a 1M-token context window in Codex for GPT-5.6 Sol. Even though we have tuned the context limit in Codex to be set optimally when it comes to performance and cost, this is a common ask, so here it is documented. A larger context window lets Codex retain more code, tool output, and conversation history before summarizing older material. You need a model that supports it. And GPT-5.6 Sol, for example, has a documented 1,050,000-token window. Open ~/.codex/config.toml and add or update these settings at the top level, before any [section] headers: ``` model = "gpt-5.6-sol" model_context_window = 1000000 model_auto_compact_token_limit = 900000 ``` The first setting selects the model. The second tells Codex to use a one-million-token context budget. The third starts automatic history compaction around 900,000 tokens, leaving some headroom. Restart Codex client and start a new session after saving. To try the configuration for a single CLI session without changing your defaults: ``` codex -m gpt-5.6-sol \ -c model_context_window=1000000 \ -c model_auto_compact_token_limit=900000 ``` Have fun, but also know that we tuned the default carefully!
21
Replying to @jerryblount @OpenAI
Ok...if you say so Codex.
24
Replying to @jerryblount @OpenAI
...I think I'm certainly missing something, I find it hard to believe, @OpenAI would toss up this alley-oop to @AnthropicAI (screenshot, my max account). $100/m on Codex gets me 258K context window, on Claude Code 1M.
1
1
37
Today I built this voice agent while on a call with it Introducing /this-needs-a-call - it's a slash command to start a phone call - grill or be grilled by the voice agent in realtime - your coding agent listens in on the call via MCP and works while you talk for hours
Thereโ€™s something magical about working on a mobile app from within the app itself
18
10
3
203
32,869
Jerry Blount retweeted
New on @digg: a GitHub tab that scores every repo for breakout potential and novelty. Plus top starred, top active, and a raw activity feed. Shipped this Friday afternoon.
14
3
3
167
100,988
Jerry Blount retweeted
Introducing SpacetimeDB 2.0. Web development at the speed of light. Come learn what fast really means.
106
102
91
1,265
803,896
Jerry Blount retweeted
No one asked for this (literally no one), but you can now use @figma-like layers in wiretext.app/
13
9
2
93
9,135
After 2-3 compactions, Warp agent just crashes, doesn't remember any of the work it did, literally have to start all the way over after upwards of an hour plus of development...SMH @warpdotdev you can do better.
1
1
88
@warpdotdev the application needs a transcript / memory feature that's native, can store user and agent turns and terminal inputs and outputs, store it all in a .txt file, accessible through a sub agent so the main context window is not used.
1
1
60
Whatโ€™s the deal with saying a store is too full to let customers in when clearly not the case @tecovas a bit too cold to be playing the โ€œoh look thereโ€™s a lineโ€ thing. What gives? Not a great first experience.
1
1
116
Cursor integration with Gemini 3 is completely broken...smh. Disappointing that @cursor releases tend to send the product into a tailspin. @warpdotdev @antigravity
2
150
We just released an open source framework that sets up agentic search and RAG in a full web UI on your own data in just two terminal commands. Meet Elysia - a decision tree based agentic system that dynamically displays data, learns from user feedback, and chunks documents on-demand. Most AI chatbots are stuck in a text-in, text-out world. But what if your AI could dynamically decide not just what to say, but ๐—ต๐—ผ๐˜„ ๐˜๐—ผ ๐˜€๐—ต๐—ผ๐˜„ ๐—ถ๐˜? Elysia is designed to completely rethink how we interact with data through AI. Three features that set it apart: ๐ŸŒณ ๐——๐—ฒ๐—ฐ๐—ถ๐˜€๐—ถ๐—ผ๐—ป ๐—ง๐—ฟ๐—ฒ๐—ฒ ๐—”๐—ฟ๐—ฐ๐—ต๐—ถ๐˜๐—ฒ๐—ฐ๐˜๐˜‚๐—ฟ๐—ฒ: Unlike simple agentic platforms, Elysia uses a pre-defined web of nodes, each orchestrated by a decision agent that evaluates its environment and strategizes the best tool to use. You can watch the entire decision process in real-time. ๐ŸŽจ ๐——๐˜†๐—ป๐—ฎ๐—บ๐—ถ๐—ฐ ๐——๐—ถ๐˜€๐—ฝ๐—น๐—ฎ๐˜† ๐—™๐—ผ๐—ฟ๐—บ๐—ฎ๐˜๐˜€: Elysia can choose from 7 different ways to present your data - tables, product cards, tickets, conversations, documents, charts, and more. It automatically analyzes your data structure and picks the most appropriate format. ๐Ÿง  ๐—”๐˜‚๐˜๐—ผ๐—บ๐—ฎ๐˜๐—ถ๐—ฐ ๐——๐—ฎ๐˜๐—ฎ ๐—˜๐˜…๐—ฝ๐—ฒ๐—ฟ๐˜: Connect your Weaviate Cloud instance and Elysia becomes an instant expert on your data. It analyzes collections, generates metadata, and understands your data structure before performing any queries. Getting started is ridiculously simple: pip install elysia-ai elysia start That's it. You get a full web interface AND a Python library. Bonus features we're excited about: โ€ข Chunk-on-demand: No more pre-chunking everything โ€ข Feedback system that learns from your preferences โ€ข Multi-model strategy for cost optimization โ€ข Everything runs from a single Python package We're already using Elysia to power the chat interface in our AI skincare app, Glowe. The entire project is open source and designed with customization in mind. Get started: ๐Ÿ”— GitHub: github.com/weaviate/elysia ๐Ÿ“– Demo: elysia.weaviate.io ๐ŸŽฎ Blog post: weaviate.io/blog/elysia-agenโ€ฆ
7
58
4
324
49,822
Replying to @SGRodriques
Check out our platform at platform.futurehouse.org, and read more at our blog post below! futurehouse.org/research-annโ€ฆ
2
16
1
62
7,305
Verba 1.0 is finally here! ๐Ÿ• With our latest release, you can now run a state-of-the-art Retrieval Augmented Generation (RAG) application locally on your computer, thanks to the new integration with @ollama. This allows you to use fantastic open source models like Llama 3, @MistralAI's 7B, or Command R from @cohere. Check out our new video featuring @aestheticedwar1, @philipvollet, and @victorialslocum, where they explain everything you need to know about setting up a full advanced RAG application in just a few minutes. Theyโ€™ll also show you the major UI upgrades we've made and go through some example use cases for handling sensitive data RAG. Watch the full video here: youtube.com/watch?v=swKKRdLBโ€ฆ
3
24
3
153
16,073
Jerry Blount retweeted
Introducing play.ai โ€” a new conversational Voice AI platform and API for creating human-like voice agents and voice interfaces. Powered by our new Large Dialogue Model (LDM), Play LDM 1.0 makes the speech generation contextual; handles turn-taking, interruption, voice energy and emotion modulation for a natural, fluid, human conversations in real-time.
32
51
26
343
105,701
Jerry Blount retweeted
We've just launched fine-tuning for GPT-3.5 Turbo! Fine-tuning lets you train the model on your company's data and run it at scale. Early tests have shown that fine-tuned GPT-3.5 Turbo can match or exceed GPT-4 on narrow tasks: openai.com/blog/gpt-3-5-turbโ€ฆ
399
1,557
505
7,216
3,528,264