AI Intelligence Archive

Language models typically construct attention values from contextual hidden states, even when some of their content may be reusable across contexts. We investigate whether token-indexed memory can replace the dedicated value projection whe…

Fine-tuning large language models on parallel data improves translation quality but can cause catastrophic forgetting. Mitigation methods are generally evaluated by retention on general benchmarks. We ask whether these findings transfer to…

Score matching has driven major advances in classical generative learning by enabling models to learn from data without evaluating intractable normalization constants, or partition functions. Yet, extending this principle to quantum learni…

OpenAI says its products now reach more than 1 billion weekly active users and 2.5 million...

Reinforcement learning with verifiable rewards (RLVR) supervises mathematical reasoning through final-answer correctness, but provides little guidance on individual tokens. On-policy distillation (OPD) supplies dense feedback on student-ge…

A pull request stalled on a quiet screenshot from an agent session that looked like performance...

A patch landed with every assertion green. The return values matched the golden file. Then a latency...

An agent's egress allowlist permits the artifact registry because builds need it. That single permitted destination is enough to run a covert...

A suspicious transaction starts an investigation. The harder part is deciding what evidence to...

Why supervisor LLM agents create infinite retry loops and silent token inflation, and how to replace hierarchical supervisor loops with deterministic typed state transitions.

A few weeks ago I promised a technical write-up. Here it is. The two questions I get most are: how...

Everyone is obsessed with making AI remember more. Nobody is working on making it forget. And that's a serious technical problem nobody talks about.

AI observability is the practice of tracking what an AI system did and why. That means looking...

We have officially entered the era of vibe coding. With AI-assisted editors and terminal agents like...

In the last article we built a working chat endpoint. Send a message, get a reply. It felt like...

Disco Doodle is my hobby site — free, no-signup drawing prompt generators, a division of my little...

I had an idea, more of a stress-test really. I have been building my IDE for a few months now, it's...

This is a submission for the Sanity Challenge, Path Two: Vibe-Code Something Strange What I...

Claude Code writes every session to disk as JSONL: one line per message, tool call and result, and a...

I spent a week adding per-agent cost tracing to a multi-agent AWS Bedrock and Strands crew. It returned a perfect answer, showed 200 OK, and still billed ~1.4x. Here is how to catch that silent waste, read-only and at $0.

I noticed it first in a Slack channel, of all places. A coworker dropped a link to some new "agent...

Opus 5.5 shipped on September 22, and its replies read differently: answer first, less...

This is the experience I want reading with Luminary to feel like — not a feature list, but a day in...

Recall@1 went from 13% to 50%. Recall@5 from 13% to 80%. Over the same week, the number of questions...

I assumed that "extract the text from this PDF" gives you the same characters no matter which library...

The plan as an editable artifact: stable task ids, dependencies as data, and the edits a model is not allowed to make.

The European Commission is proposing a common rating scheme for data centres intended to improve...

Most "AI support agent" tutorials end with a chat box that confidently makes something up. The fix is...

Data assimilation estimates a dynamical state from partial and noisy observations. Classical ensemble filters are efficient but restrict analysis updates through finite sample covariance and affine Gaussian distribution. We introduce the F…

Machine-learned interatomic potentials (MLIPs) for solid-liquid interfaces in advanced materials applications, e.g., electrochemistry, catalysis and corrosion, require training data that samples both liquid environments, the solid and the …

Most Turkish-capable large language models (LLMs) are evaluated using general-purpose benchmarks rather than long, structurally complex domain documents. This paper evaluates five open-weight 7B-8B models for Turkish document question answ…

Decoder-only Transformer language models cache keys and values (KV) to reuse past computation during generation. Sharing KV across layers saves storage but reduces the diversity of representations available across depth. We study what loca…

Effective summarization of interrogative dialogues is a critical task in forensic and investigative settings, requiring high factual accuracy, coherence, and attribute-specific relevance. In this work, we introduce CASPER, a novel Chain-of…