Agents Made Simple
I have built many AI agents, and all frameworks felt so bloated, slow, and unpredictable. Therefore, I hacked together a minimal library that works with JSON/dict/kwargs definitions for each step, allowing you a simpler way to define reproducible agents. It supports concurrency for up to 1000 calls/min, giving you speed and predictability in your workflows. Install pip install flashlearn Input is a list of dictionaries Simply take user inputs, API responses, and calculations from other tools and feed them to FlashLearn. user_inputs = [{"query": "When was python launched?"}]…
What it does
In the maker’s words, at launch
I have built many AI agents, and all frameworks felt so bloated, slow, and unpredictable. Therefore, I hacked together a minimal library that works with JSON/dict/kwargs definitions for each step, allowing you a simpler way to define reproducible agents. It supports concurrency for up to 1000 calls/min, giving you speed and predictability in your workflows. Install pip install flashlearn Input is a list of dictionaries Simply take user inputs, API responses, and calculations from other tools and feed them to FlashLearn. user_inputs = [{"query": "When was python launched?"}] Skill is just a simple dictionary A skill is an LLM’s ability to perform a task, containing all the necessary information. You can create your own, use predefined samples, or generate them automatically from example data. ConvertToGoogleQueries = { "skill_class": "GeneralSkill", "system_prompt": "Exactly populate the provided function definition", "function_definition": { "type": "function", "function": { "name": "ConvertToGoogleQueries", "description": "Convert the given question into between 1 and n google queries to answer the given question.", "strict": True, "parameters": { "type": "object", "properties": { "google_queries": { "type": "array", "items": {"type": "string"} } }, "required": ["google_queries"], "additionalProperties": False } } } } Run in 3 lines of code Load the skill, create tasks (a list of dictionaries), and run them in parallel. Results are easy to parse in downstream steps. skill = GeneralSkill.load_skill(ConvertToGoogleQueries) tasks = skill.create_tasks([{"query": "User's query"}]) results = skill.run_tasks_in_parallel(tasks) Get structured results The output is a dictionary, where each key corresponds to an index in the original list. This lets you keep track of results easily. flash_results = {'0': {'google_queries': ["QUERY_1", "QUERY_2"]}} Pass on to downstream tasks Use the structured JSON output in your next steps. queries = flash_results["0"]["google_queries"] results = SimpleGoogleSearch(GOOGLE_API_KEY, GOOGLE_CSE_ID).search(queries) msgs = [ {"role": "system", "content": "insert links from search results in response to quote it"}, {"role": "user", "content": str(results)}, {"role": "user", "content": "When was python launched?"} ] print(client.chat.completions.create(model=MODEL_NAME, messages=msgs).choices[0].message.content) Feel free to ask anything!
Does the same job
all alternatives →- ACAI code reviewer with Python agent called from Rust, streaming works2025 · medium.com · ▲6
Built an AI code reviewer using Letta (Python) that I can call natively from Rust applications. The interesting part: real-time streaming works perfectly across the language boundary with zero hassle using RunAgent. The agent runs in Python with persistent memory, leverages the best in house agentic memory management with Letta (Pythonic AI agent framework), and my rust code just uses it (kinda) natively, though Letta has no Rust bindings. And, streaming works like magic. No FFI, no complex bridges - just native async/streaming that feels like calling any Rust librar, but without…
- RRReka Research – A SOTA research agent2025 · reka.ai · ▲6
Hi HN, We built something new! It is an AI agent that can: - iterate through up to a dozen sources (web or GDrive) - reason between hops in natural language (exposed in the trace) - generate structured JSON so downstream code can consume results deterministically It not only delivers state-of-the-art performance on both SimpleQA and Reka Research-Eval, but also shatters the notion that cutting-edge AI must come at a premium, being significantly more affordable than alternatives. Tech stack highlights - Base model: Reka Flash 3.1 (trained from scratch, post-trained with RL on verifiable…
- LALLM agents that write Python to analyze execution traces at scaleMar 2026 · github.com · ▲5
We combined Stanford's ACE (agents learning from execution feedback) with the Reflective Language Model pattern. Instead of reading traces in a single pass, an LLM writes and runs Python in a sandbox to programmatically explore them - finding cross-trace patterns that single-pass analysis misses. The framework achieved 2x consistency improvement on τ2-bench.
- UOUpdates on Burr (OS) – a full-stack AI agent framework2024 · burr.dagworks.io · ▲8
Hey HN, In the months since we initially released Burr (https://news.ycombinator.com/item?id=39917364), we have been hard at work. We wanted to share some of the most exciting changes we’ve made to build Burr out as a full-stack development framework for AI agents. In case you don’t recall, Burr is an open-source python library that makes it easier to build and debug GenAI applications & agents by representing them as graphs of simple python objects/functions. Burr only abstracts away system-level concerns (state persistence, debugging, observability), and does not…
- HAHyperFlow – A self-improving agent framework built on LangGraphApr 2026 · ▲8
Hi HN, I am Umer. I recently built an experimental framework called HyperFlow to explore the idea of self-improving AI agents. Usually, when an agent fails a task, we developers step in to manually tweak the prompt or adjust the code logic. I wanted to see if an agent could automate its own improvement loop. Built on LangChain and LangGraph, HyperFlow uses two agents: - A TaskAgent that solves the domain problem. - A MetaAgent that acts as the improver. The MetaAgent looks at the TaskAgent's evaluation logs, rewrites the underlying Python code, tools, and prompt files, and then tests the new…
- IMI made AI agent lib that you will understand2025 · github.com · ▲7
Every time I wanted to use LLMs in my existing pipelines the integration was very bloated, complex, and too slow. This is why I created a lightweight library that works just like scikit-learn, the flow generally follows a pipeline-like structure where you “fit” (learn) a skill from sample data or an instruction set, then “predict” (apply the skill) to new data, returning structured results. High-Level Concept Flow Your Data --> Load Skill / Learn Skill --> Create Tasks --> Run Tasks --> Structured Results --> Downstream Steps And the bast part: Every step can be saved and reused as…
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 16d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 26d ago · cactuscompute.com
Launched alongside, January 2025
the whole month →- IM
Hello! I'm Byran. I spent the past ~6 months engineering a laptop from scratch. It's fully open-source on GH at: https://github.com/Hello9999901/laptop
Dev tools · 2025 · byran.ee
- TITetris in a PDF▲1,289
I realized that the PDF engines of modern desktop browsers (PDFium and PDF.js) support JavaScript with enough I/O primitives to make a basic game like Tetris. It was a bit tricky to find a union of features that work in both engines, but in the end it turns out that showing/hiding annotation "fields" works well to make monochrome pixels, and keyboard input can be achieved by typing in a text input box. All in all it's quite janky but a nice reminder of how general purpose PDF scripting can be. The linked PDF is all ASCII so you can just open it in a text editor, or have a look at…
Life & fun · 2025 · th0mas.nl
Create lifelike, personalized AI avatars from text prompts
AI · 2025 · jogg.ai
