research + solutions
Cohort 02 of the Research Lab Programme is now closed. Register your interest for the next cohort — we’ll get in touch when applications open.
Cohort 02 · applications closed
Register interestThe next cohort has the focus of teaching machines to think and interpret as we do. Share a project or research idea that aligns with the topics below.
Next cohort · register interest
This is for curious, early-career researchers and builders — students, junior engineers, data/ML enthusiasts, designers, and creatively technical people.
Read the TEV Research Engine white paper and the pitch deck if you want context for a project or research idea.
Leave your name, email, and a project or research idea here, or email info@asycd.online. We’ll follow up when the next cohort opens.
Notes and studies from the lab archive — the same trail as the research lab.
13 October 2026
Classification is the same. Kinship tags now share a room, so Sad can meet Melancholic. Meaning returns to 1× so we can isolate the effect.
Read A Letter on Scoring Changes — 13 October 20269 October 2026
Cohort 02 has started a 12-week programme on teaching machines to interpret as we do. The TEV Research Engine is in beta: retrieve by meaning, not keywords.
Read Introducing Cohort 02 — Teaching Machines to Interpret as We Do with TEV21 August 2026
Applications are open now for Research Lab Cohort 02 — teaching machines to navigate the internet as a living thematic landscape.
Read The Research Lab is Back — Cohort 02 Applications Open Now12 March 2026
The Asycd research lab officially launched the week of February 16th, initiating a high-intensity immersion into the current state-of-the-art in Generative AI. Learn about the first two weeks of the lab and how our researchers are doing.
Read Asycd AI Research (March 2026): Demos in Multi-Agent Orchestration and Bayesian Prompt Compression28 March 2026
In this week’s task, we tasked the researchers with creating architecture diagrams to visualize their solutions. An architecture diagram is visual representation of the key components, processes, inputs, and outputs involved in a system, and how they interact end-to-end. From the moment you press ru
Read How to Think About Software Architecture Design? - Week 5 Research Lab Update (Mid March 2026)14 March 2026
The Asycd Research Lab explores cutting-edge AI, automation, and prompt engineering - turning experimental ideas into practical tools, workflows, and potentially commercial products.
Read introducing - “the research lab”8 December 2025
TEV2 leverages various different web data sources to generate enriched, multi-perspective AI art, blending agentic creativity with scalable production for unique visual experiences.
Read TEV2 and Agentic Creativity: Exploring the World Through a Single Frame7 December 2025
AI agent long-term memory implementation guide: Redis vector database, conversation history storage, persistent context, sliding window architecture.
Read How to Configure Long-Term Memory in AI Agents: A Practical Guide to Persistent Context29 November 2025
Ever wondered how punctuation and word placement shape AI-generated responses? In this article, we explore the hidden mechanics behind prompt engineering, revealing how subtle tweaks—like commas and phrase order—can dramatically alter AI outputs. Our latest research dives into practical demonstrat
Read Visualizing the Effects of Prompt Reordering and Punctuation in AI Prompts using AI Art29 November 2025
Inference-only compute is revolutionizing AI economics by letting users pay only for the compute used during inference, eliminating costs for idle resources. Platforms like Modal dynamically allocate GPUs for precise, real-time usage, enabling startups to experiment freely, scale efficiently, and
Read A New Era for AI Costs: The Innovation of Inference-Only Compute Costs29 November 2025
Discover how timestamped embeddings are revolutionizing Retrieval-Augmented Generation (RAG) by enhancing AI systems with temporal awareness. Learn about their crucial applications in finance, technology, and more, and explore Python implementation techniques to boost data relevance and retrieval ac
Read Timestamped Embeddings for Time-Aware Retrieval-Augmented Generation (RAG)29 November 2025
A truly intelligent AI will be able to perform multiple tasks with ease and accuracy. No need for separate writing assistants, customer support etc. We propose an agentic and dynamic system prompt approach that uses a separate LLM to analyse the users queries and transform the insights into a usefu
Read Enhancing LLM Adaptability Through Dynamic Prompt Engineering29 November 2025
The filter bubble paradox highlights how AI-driven personalization enhances user experience while limiting exposure to diverse viewpoints, risking creativity and broadening inequality in digital content.
Read the filter-bubble problem in recommendation algorithms + the potential impact of generative ai29 November 2025
The TEV1 algorithm transforms regular images prompts into eye-bending visuals using generative AI and combinatorics.
Read The Beauty of Systems— TEV1 Algorithm Breakdown29 November 2025
Refine your AI prompts with Asycd
Read 10 Tips for Improving Your AI Prompts — LLM Prompt Engineering29 November 2025
The RAG approach enhances an LLM
Read Advanced Multi-Query RAG Techniques for Better Chatbot Experiences29 November 2025
Discover the power of ReAct prompting for AI responses. Our article delves into the ReAct framework, enhancing decision-making in language models with human-like reasoning1. Learn how ReAct surpasses Chain-of-Thought, offering structured, precise AI outputs.
Read ReAct Prompt Framework: Enhancing AI’s Decision Making with Human-Like Reasoning19 July 2025
Cross-encoders offer a powerful lens into the semantics of dialogue. By jointly analyzing user queries and conversation history, they surpass basic similarity methods to reason about intent, entailment, and contextual nuance. In our work at Asycd, they serve as the final arbiter for query-to-conte
Read Cross-Encoder Models for Efficient Query-to-Context MappingStudy
How embeddings and context loading trade off when you need relevance without drowning the model.
Open study: Embeddings and context loading studyStudy
Chunking strategies that adapt to structure — so retrieval stays useful as documents grow.
Open study: Dynamic text chunking using LLMsCohort 01 · Feb–May 2026
Three researchers. Three tracks. The inaugural sprint finished in late May 2026. Cohort 02 applications are now closed.
01
Bayesian Prompt Compression via Structural Pattern Discovery
02
Distributed State Consistency & Cost-Aware Orchestration for Multi-Agent Systems
03
Emergent Memory and Continual Reasoning in Generative AI Agents via Biologically Inspired Learning Dynamics
Outputs from Cohort 01: demos, packages, and systems that survive contact with real-world prompts.
Joel-Allen Calliste
Bayesian prompt compression via Structural Pattern Discovery — a Gaussian process that shrinks prompts while holding performance. Installable as prompt-compress; demo site forthcoming.
pip install prompt-compress
Saksham Kapoor
Distributed state consistency and cost-aware orchestration for multi-agent systems — addressing agentic entropy, context bloat, and FinOps routing.
The Asycd research lab gives early-career generative AI enthusiasts, undergraduates, and indie programmers a chance to contribute to meaningful prompt engineering and applied AI research.
We have researched prompt engineering since 2022 and are scaling those efforts while identifying high-potential people and ideas. The lab runs in 3-month sprint periods with a small cohort each cycle.
Browse research notesCurrent status: Cohort 01 finished in late May 2026. Cohort 02 applications are now closed — register interest for the next cohort.
Follow outputs on the blog — literature reviews, V0 demos, architecture notes, and shipped packages from the first sprint.
