GoDataAnalytics @GoDataAnalytic

Andrej Karpathy: "Prompting is going away. Delete everything, keep Graph." In just 69 minutes, he shows how to build graphs and why they're the only layer that survives in the end Prompts → Agents → Loops → Graphs a loop keeps one agent running until the job is done a graph decides which agents exist and what each one passes to the next anyone can build an agent, almost nobody builds the graph that runs them watch it today, then save the full guide on graphs below before everyone starts catching up ↓

Mori @0xjmori

a day ago

Claude Opus 5.5 could turn a single AI bot into an entire motion design studio. It can direct scenes, animate product interfaces, render videos, catch visual mistakes and deliver polished results. In this article, I show you how. x.com/i/article/2111…

30 23 247 389K 755

This is Pele, my new rendering library for diagrams and charts, compatible with the latest Mermaid syntax. - 100-500× faster than mermaid.js - 6× smaller bundle - No dependencies - Easy to theme - Responsive to device width In pursuit of making Obsidian insanely fast and cohesively designed, I always found the Mermaid library to be too big, too slow, and too hard to theme. Pele will be the new renderer in the next version of Obsidian, but you can start using it now. Pele [ˈpɛlɛ] is named after the goddess of volcanoes and fire. It's open source under the MIT license. Try the playground!

keyboard_arrow_left Previous keyboard_arrow_right Next

Chess, played as doodles on notebook paper. Made with Sonnet 5.5.

Six projects from the last few weeks that caught our attention: A wildfire analysis prototype that maps the fire’s spread over time and tracks hot spots, made with Opus 5.5.

Claude writes about 90% of Claude Code's code. From June to September, merged PRs more than doubled while the share reverted fell by more than 40%. We wrote about how we changed code review to get there, and how to adopt it: claude.dev/blog/solving-c…

This is f*cking insane. Chinese researchers just dropped a paper with a title that will make every agent framework nervous: "The End of Harness Engineering". The argument fits in one line: writing the program around the model is finished as a skill. A harness written by a human is one person's guess about one task on one afternoon. It ships, nobody re-reads it, nobody scores it, and the agent runs that guess for months. A harness written by the loop is a file that grades itself. A small 4B model reads what failed, writes 8 edits, and only the edit that scores best on held-out questions survives. The big model never changes. The file around it does, every round. So the human job collapses to one artifact: the scoring set. You write the questions, never the program. Frameworks turn disposable. The exam turns into the job. That is the exact loop I built and documented below: one harness.py, 300 Kimi K3 workers, 37 rewrites in 41 rounds, zero opened by me. The whole paper in one line: write the exam before the code ↓

slash1s @slash1sol

2 months ago

x.com/i/article/2092…

25 14 187 658K 417

🎯How the Claude Code team does code review🎯 claude.dev/blog/solving-c… A must-read from Anthropic’s Claude Code team on how code review evolves for agent-generated code: multi-agent verification, risk-based approvals, continuous feedback, and fast rollback.

THIS IS A PURE F*CKING GOLDMINE Anthropic just shared the sharpest trick for Claude Code The idea: let Opus 5.5 write the plan as a script, and swarm the new Haiku 5.5 in parallel Huge sweeps with zero Opus burn: the script holds the loop and Opus only sees the final diff You just launch Claude Code like this: claude --model opus --effort ultracode here's the split: → Opus 5.5 turns your objective into an executable workflow script → Haiku 5.5 swarms every file in parallel at 340 tok/s to extract candidates → Sonnet 5.5 patches each fix and cross-checks findings in its own isolated copy → you stop burning Opus context on a hundred agent transcripts Opus only shows up at three moments: → authoring the workflow script before the run → reading verified findings after the swarm finishes → signing off before anything ships to git Opus plans · Haiku scans · Sonnet ships JEV engineering is the same move one layer down: the forks that need no thinker (which file, which tool, retry or stop) go to JEV in milliseconds, and the big models only see the ones that split the full tree > Opus 5.5 runs the main session and writes workflow.js > pipeline() fans out the new Haiku 5.5 to audit every file > parallel() schema validators drop false positives in V8 > Sonnet 5.5 patches each fix in an isolated worktree > verify cross-checks all diffs before reporting > Opus 5.5 reads only what survived verification > JEV routes micro-forks in milliseconds paste the tree and this prompt into Claude Code ↓ "Rebuild my Claude Code setup around this tree: 1. Run the main session on Opus 5.5 with ultracode on 2. In every workflow, run read-only scan stages on the new Haiku 5.5 and edit stages on Sonnet 5.5, each edit in its own isolated copy 3. Add one rule to CLAUDE.md: every finding gets an independent verifier agent before it's reported, and fix loops stop after two rounds with no progress 4. Route fast deterministic forks (tool choice, path selection, retries) through JEV Show me every change as a diff first, no edits until I say go" ↳ code.claude.com/docs/en/workfl… save it, it'll come in handy

mirku @mirku21

a month ago

x.com/i/article/2098…

25 68 617 1.9M 2K

Google just open-sourced its Harness² It's a plug-and-play framework for recursive agent harnessing in an open world. Agents can perform, and learn from, tasks its harness was never designed for. Built around 8 components: 1. System prompt 2. Guardrails 3. Memory 4. Tool guidance 5. Subagents 6. Skills 7. Scripts 8. Plugins Link: github.com/google-researc…

We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks: Writing. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent. But even better: Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better: Web pages. Ask for output "in HTML" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better: Explainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like "Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work! In summary: - As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding. - Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.

Today we’re introducing Gemini 4 Argon. It delivers frontier performance in complex workflows across real-world software engineering, knowledge work, and cybersecurity defense with an industry-leading 1M token output limit.

Jev for RAG, clearly explained! Hybrid search gives you a shortlist. It does not decide which passages contain evidence, which are merely adjacent, or whether the evidence is strong enough to answer. That missing judgment is where Jev fits. Jev sits between retrieval and generation. It does not replace BM25, embeddings, or the LLM. It evaluates the candidates they produce before those candidates enter the context window. → Retrieve wide Combine dense and keyword search, then merge the results with reciprocal rank fusion. This gives Jev a broad candidate pool, such as the top 20 passages shown in the visual. Retrieval still sets the ceiling. If the right passage is missing from this shortlist, Jev cannot recover it. → Judge every candidate together Send the query as Jev’s state. For each candidate, ask a typed yes-or-no question such as “Does C7 help answer this query?” Jev evaluates every question in one packed request and returns a calibrated probability for each passage. This avoids making a separate model call for every query-passage pair. → Let code apply the threshold Your application compares each probability with a threshold. Candidates above it continue to the LLM. Everything below it is removed before generation. Jev makes the fuzzy judgment. Code remains responsible for the actual decision. → Gate the entire answer The same request can check whether the retained passages make the query answerable. It can also flag signs of prompt injection inside a candidate. If answerability falls below the threshold, the application can skip the LLM and return “not in the documents.” The injection score should remain a filtering signal, not a security boundary. The result is a cleaner division of work. Hybrid search retrieves broadly. Jev reranks, filters, and decides whether sufficient evidence exists. The LLM writes only from the passages that survive. Jev’s value here is not simply moving passages up or down a list. It turns relevance into an explicit probability that your application can inspect, threshold, and act on. To summarise: - Retrieval finds the candidates. - Jev decides what deserves context. - The LLM writes the grounded answer. ---- I also built an open-source project showing how to use Jev as a judge for AI observability with Comet Opik. It evaluates support traces for groundedness, request coverage, action honesty, and helpfulness, then records the results as an auditable experiment. You can explore the project here: github.com/patchy631/jev-… My article on how Jev works is quoted below.

Akshay 🚀 @akshay_pachaar

a month ago

x.com/i/article/2100…

131 664 6K 2.5M 11K

Lots of discussion out there about our next model(!), so I wanted to give an early look as soon as possible. Introducing Gemini 4 Argon! It shows frontier performance in complex workflows, cyber defense and software engineering. Teams are using it extensively at Google, from coding to quantum computing, great feedback. Here’s a look at the benchmarks:

Took a while, but here it is, a mega write-up on language models for text classification and, of course, Jev. It's basically a visual guide to RNNs, CNNs, transformers, and calibration, with hands-on experiments on accuracy and efficiency.

muse is shopping for me. i'm still not getting clients. gave Explee the link. it finds the buyers and writes the emails. the errand agent was never going to sell. $30 free 👇 explee.com/auto-gtm/x/tw-…

Introducing Claude Sonnet 5.5, the second model in the Claude 5.5 family. It’s a clear upgrade over Sonnet 5, runs more than 30% faster, and costs up to 30% less for most work.

@typesafeai How's RLCD rewarding calibration, and what does RL add over post-hoc scaling?

🚨 لقيت حاجة من Cloudflare تستحق التجربة فعلًا. أطلقوا Security Audit Skill مفتوحة المصدر، وبتخلي وكيل الذكاء الاصطناعي يقدر يعمل تدقيق أمني بشكل منظم، بدل ما تكون العملية مجرد بحث عشوائي عن الثغرات. 🔍 الـSkill بتقسم التدقيق لـ6 مراحل: 🔹 فهم بنية النظام ونقاط الهجوم 🔹 البحث عن الثغرات بشكل شامل 🔹 التحقق من كل ثغرة بشكل مستقل 🔹 تنظيم النتائج 🔹 مراجعة نهائية بواسطة وكلاء مختلفين 🔹 إنشاء تقرير أمني تلقائيًا 📄 لكن أكتر حاجة عجبتني هي فكرة التحقق 👀 الوكيل اللي بيكتشف الثغرة مش هو نفسه اللي بيتأكد إنها حقيقية. وده بيساعد على تقليل الـFalse Positives، بدل ما الوكيل يكتشف مشكلة ويعتبرها ثغرة وخلاص. والأداة مفتوحة المصدر، مجانية، ومرخصة برخصة MIT. ⚡ لو بتبني وكلاء ذكاء اصطناعي أو مهتم بالأمن السيبراني، دي من الحاجات اللي تستاهل التجربة. الرابط 👇

‏عادل | مبرمج @AdelDeveloperX

2 months ago

x.com/i/article/2091…

24 80 464 4.9M 917

Locally Laya just beat cloud Jev at live Tetris. Laya System 1 model made Decision time: 40–50 ms. on a 16GB MacBook Air. Jev, running in the cloud, sat above 300 ms. That is enough for 60+ decisions per second.

Md Ismail Šojal 🕷️ @0x0SojalSec

4 weeks ago

Laya is the open-source Jev that already beats close Jev on latency and calibration. - 10x faster than TypeSafe Jev on the published numbers. - Laya is multilingual, and already ahead of Jev on speed. - 38ms single query vs 400ms Locally. - No per-token bill. - A System-1

19 52 442 81K 460

If you're an AI Engineer, here are 5 projects you must build. These projects will force you to make the same decisions production teams make every day. 1️⃣ 𝗕𝘂𝗶𝗹𝗱 𝗮 𝗥𝗔𝗚 𝗣𝗶𝗽𝗲𝗹𝗶𝗻𝗲 𝗮𝗻𝗱 𝗶𝗻𝘀𝘁𝗿𝘂𝗺𝗲𝗻𝘁 𝗶𝘁 Don't just wire up retrieval + generation: instrument it. Track retrieval precision, context relevance, and hallucination rate on a held-out eval set. ↳ Tools: LangChain or LlamaIndex, a vector DB, Ragas for eval ↳ YouTube: youtube.com/watch?v=XebKOY… 2️⃣ 𝗙𝗶𝗻𝗲-𝗧𝘂𝗻𝗲 𝗮 𝗦𝗺𝗮𝗹𝗹 𝗢𝗽𝗲𝗻-𝗪𝗲𝗶𝗴𝗵𝘁 𝗠𝗼𝗱𝗲𝗹 𝗳𝗼𝗿 𝗢𝗻𝗲 𝗡𝗮𝗿𝗿𝗼𝘄 𝗧𝗮𝘀𝗸 Take a 7B model, pick one job (classification, extraction, a house style), and fine-tune it with LoRA. Compare it against prompting a frontier model for the same task. ↳ Tools: Hugging Face, PEFT/LoRA, a single GPU (or Colab) ↳ YouTube: youtube.com/watch?v=F29Y_y… 3️⃣ 𝗗𝗲𝘀𝗶𝗴𝗻 𝗮𝗻𝗱 𝗥𝘂𝗻 𝗮𝗻 𝗘𝘃𝗮𝗹 𝗛𝗮𝗿𝗻𝗲𝘀𝘀 𝗳𝗼𝗿 𝗟𝗟𝗠 𝗢𝘂𝘁𝗽𝘂𝘁𝘀 Build a repeatable pipeline that scores model outputs, not vibes, actual metrics: exact match, LLM-as-judge, human-in-the-loop spot checks. ↳ Tools: Python, a judge model, a small labeled dataset ↳ YouTube: youtube.com/watch?v=rS6wZf… 4️⃣ 𝗕𝘂𝗶𝗹𝗱 𝗮 𝗠𝘂𝗹𝘁𝗶-𝗔𝗴𝗲𝗻𝘁 𝗦𝘆𝘀𝘁𝗲𝗺 𝗧𝗵𝗮𝘁 𝗛𝗮𝘀 𝘁𝗼 𝗖𝗼𝗼𝗿𝗱𝗶𝗻𝗮𝘁𝗲 Two or more agents, one shared goal, real handoffs: not a single agent calling tools in a loop. ↳ Tools: LangGraph, function calling ↳ YouTube: youtube.com/watch?v=1SFZz8… 5️⃣ 𝗦𝗵𝗶𝗽 𝗮 𝗙𝗲𝗮𝘁𝘂𝗿𝗲 𝗪𝗶𝘁𝗵 𝗚𝘂𝗮𝗿𝗱𝗿𝗮𝗶𝗹𝘀, 𝗡𝗼𝘁 𝗝𝘂𝘀𝘁 𝗮 𝗗𝗲𝗺𝗼 Add input validation, output filtering, rate limiting, and graceful fallback when the model fails or times out. ↳ Tools: Any LLM API, a FastAPI wrapper, slowapi ↳ YouTube: youtube.com/watch?v=tW3hEy… AI Engineering is retrieval, evaluation, fine-tuning, orchestration, and reliability engineering, all applied to models instead of deterministic code. → If you have GPU access, build all five and benchmark them against each other. → If you don't, most of this runs fine on a laptop with a small model or a frontier API. → Even just designing the eval harness on paper will sharpen how you think about AI systems. Pick one. Build it this weekend.