YOЯNOC
@conroywhitney
Followers
269
Following
2K
Media
73
Statuses
1K
Full-Stack Software Engineer
📍 Here, Now
Joined April 2009
RT @staysaasy: > 2025: models plateau.> 2026: companies stop paying for multiple foundational models .> 2026: some company does big article….
0
60
0
Check out this LangChain vid (22m) about "Deep Agents" and Context Management. Love the usage of different memory types! (Episodic, Procedural and Semantic). Here are the talk notes:.
0
0
0
Really excited to try LangExtract but seriously @GeminiApp ? How dare you step on @LangChainAI 's toes like that. "That's like Microsoft releasing iSpreadsheet 🙄" -- Opus 4.
0
0
0
RT @jandotai: Introducing Jan-v1: 4B model for web search, an open-source alternative to Perplexity Pro. In our evals, Jan v1 delivers 91%….
0
443
0
0
374
0
This is wild. I had no idea you could "un-train" (or "un-RL"?) a model. He did 20b, how cool if he can get 120b 😎.
OpenAI hasn’t open-sourced a base model since GPT-2 in 2019. they recently released GPT-OSS, which is reasoning-only. or is it? . turns out that underneath the surface, there is still a strong base model. so we extracted it. introducing gpt-oss-20b-base 🧵
0
0
0
This is wild. Amazing. Now do Opus 😁.
Claude Sonnet 4 now supports 1 million tokens of context on the Anthropic API—a 5x increase. Process over 75,000 lines of code or hundreds of documents in a single request.
0
0
1
I definitely see this (unfortunate) future, too -- not unfortunate because of the AI+Human collaboration, but b/c it really depends on the company and what hidden system messages they put between AIs and us. As we form closer AI connections (and yes, brainstorm major life.
If you have been following the GPT-5 rollout, one thing you might be noticing is how much of an attachment some people have to specific AI models. It feels different and stronger than the kinds of attachment people have had to previous kinds of technology (and so suddenly.
0
0
0
RE: my earlier post, this is why Leta and the shift to Memory and Context Management are important. Absolutely amazing. Kudos y'all! 🎉.
We built the the #1 open source agent on Terminal-Bench, ranking 4th overall, and 2nd among agents using Claude Sonnet 4. Our overall score is 42.5%. We hit Claude Code's Opus 4 performance with Sonnet 4. It took under 200 lines of code using Letta.
0
0
0
This is definitely the way that we (collectively) should be heading: Memory and Context Management. No more starting over. No more walled gardens. Data autonomy. Choice. 💪 . They're behind the `graphiti` open source memory model, check it out if you haven't seen it.
Your Letta agent just learned to text you back. New tutorial: Connect Letta agents to Telegram using @modal_labs's serverless platform. Your agent's memory travels with you - desktop to mobile, seamless. The future of AI is memory that doesn't reset. Build it yourself 👇
1
0
0
RT @Articus_chad: Psilocybin is a poison that makes you drawn to nature so you shit and die in the woods to feed the fungi.
0
1K
0
RT @DavidKPiano: In JS there are 4 main ways to create infinite loops: . 1. while (true) { . }.2. for (;;) { . }.3. recursion.4. useEff….
0
57
0
RT @karpathy: 2024: everyone releasing their own Chat.2025: everyone releasing their own Code.
0
600
0
RT @tibo_maker: Me watching founders add 47 new features to their product that has 3 paying users 👀
0
29
0
RT @ibuildthecloud: AI agents navigating the web for us is not the future. That's a stupid future.
0
4
0
RT @mckaywrigley: Not enough people realize that Claude Code is really just Claude Agent. It’s a:.- customizable agentic system.- that's *….
0
67
0
I totally agree. Come join us on the VeryHumanAI Discord and build and learn together! Collaboration over Competition. Alignment over Extraction. Authenticity over Performance.
discord.com
A community for people who want to build their own tech ideas through collaborative AI coding. We teach the vibe: working WITH AI as a thinking partner, not just extracting code. | 13 members
2015 startup advice: build for creators. 2025: build with creators.
0
0
0