JS Wei (Jack) Sun

Datasette ships app panel, Phonon TTS hits 28ms, Dharma's 2B tops Opus 5

Every URL the pipeline pulled into ranking for this issue — primary sources plus the supporting and contradicting findings each Researcher returned. Inline citations in the issue point back here.

← Back to the issue

Sources

datasette-agent 0.5a0 simonwillison.net

Release: datasette-agent 0.5a0 This is a significant release of Datasette Agent . The most notable new feature is called the panel - it lets the agent chat open and then sit next to pages from Datasette, for example a table or query page. When combined with datasette-apps this means you can prompt a database-backed app into existence and then modify it from the chat that sits next to that app. You can try it by signing in with your GitHub account to agent.datasette.io . Tags: projects , dataset…

datasette-llm 0.1a10 simonwillison.net

Release: datasette-llm 0.1a10 Fixes a bug that affected my agent.datasette.io demo site, where the set of available LLM models was not correctly restricted to just those in the configuration. Tags: datasette , llm

datasette-apps 0.3a0 simonwillison.net

Release: datasette-apps 0.3a0 Datasette Apps adds support for hosting and editing custom HTML applications inside of Datasette. These can be created by hand, but it’s easier to create them using LLMs via Datasette Agent . This release improves that integration: Apps can now be tagged in Datasette Agent chat messages by typing # , which autocompletes from the apps the user has permission to view. New app_add_stored_queries() and app_remove_stored_queries() tools let the agent manage which stored…

A Closer Look at Phonon TTS. Launching High-Quality On-Device Speech That Tops the Benchmarks. huggingface.co

Dharma Field Extractor: Any Document In, Structured Fields in JSON Out huggingface.co

Securing the agentic enterprise with data lineage technologyreview.com

Agent-to-agent handoffs now move data across restricted systems with no human in the loop, and a trio of pieces digs into the fallout: data lineage gaps, degraded analytics quality, and runaway token economics where one coding task burns 10-50x a chat prompt.

Raising the quality of insights in agentic analytics technologyreview.com

In just a few years, generative AI has changed the face of enterprise analytics, notably by making analytics accessible to users across the organization. Agentic AI promises to do even more, by automating entire workflows and taking actions based on agent insights. But there is a problem: The quality of insights that agents currently generate…

Stop counting tokens, start counting what ships technologyreview.com

When a developer gives an AI agent a coding task, the agent does more than answer once. It plans, generates code, runs tests, and repeats the process to correct errors. A single task can use 10 to 50 times more tokens than a basic chat prompt. The more useful the agent becomes, the more calls…

llm 0.37.1 simonwillison.net

The CLI tool for talking to language models gets a patch release pulling in fixes from Steven Huang, Tao An, Arjan Mossel, and XQ. No new features — purely stability work on top of the 0.37 line.

References

LayerLens — The Hidden Flaws of AI Benchmarks layerlens.ai

self-reported scores are, on average, 6.4% higher than those obtained by third-party testers

Contrary Research — AI Benchmarks Climbing the Wrong Hills research.contrary.com

labs grade their own homework, creating a conflict of interest that the industry is now trying to replace with secret test sets and proctored examinations

Minerka — Dharma Field Extractor writeup minerka.com

evaluations should focus on hard cases — handwriting, official stamps, skewed or low-quality scans — and on multi-page tables where schema consistency across page breaks is a known challenge for small-parameter models

Dharma-AI corporate site dharma-ai.com.br

Dharma-AI is a Rio de Janeiro spin-off from EloGroup pursuing a ‘specialization over scale’ strategy with 2B–7B SLMs, backed by selection into the AWS Global AI Accelerator and NVIDIA Inception

Particula.tech — Why Specialized 7B Models Outperform GPT-5 particula.tech

a 7B model fine-tuned on 10,000 domain-specific examples can achieve 99.8% JSON validity and over 96% classification accuracy, whereas general models like GPT-5 or Claude Opus may hover around 85% validity

local-ai-zone — Best AI OCR Models 2026 Ranking local-ai-zone.github.io

the 0.9B PaddleOCR-VL-1.6 has reportedly outperformed GPT-5.2 on OmniDocBench v1.6, achieving 96.34% extraction while running at a fraction of frontier compute cost

daily.dev summary of datasette-apps daily.dev

datasette-apps is a plugin designed to host custom HTML/JavaScript applications directly inside a Datasette instance, granting them unique read-write capabilities via SQL queries and stored canned queries… it provides a copy-paste prompt generator that feeds a database’s schema and sandbox rules into external LLMs to write the app code.

agentpatterns.ai — Browser sandbox for agent-generated HTML agentpatterns.ai

The core of the security model is a tightly constrained sandbox using sandbox=“allow-scripts allow-forms”… intentionally omits allow-same-origin… CSP injected via a tag in srcdoc with default-src ‘none’ disables all outbound HTTP requests… Communication occurs via the Channel Messaging API rather than standard window.postMessage.

thewaynext.com — Claude Artifacts vs ChatGPT Canvas thewaynext.com

Claude Artifacts is frequently praised for its ‘magic trick’ immediacy — rendering apps as they are typed — but is criticized for performing full code rewrites for minor changes… datasette-apps is positioned as a tool for building durable, data-integrated internal tools.

n1n.ai — Exploring Datasette Agent for LLM data analysis explore.n1n.ai

Internal research using DSPy has identified that models sometimes guess column names incorrectly if the schema listing in the prompt is too sparse… benchmarks for SQL agents indicate a ceiling at roughly 85% accuracy, with failure rates surging as high as 66% in production schemas plagued by schema mix-ups.

colrows.com — Vanna.ai alternatives colrows.com

Fine-tuning models or adding a semantic layer (like those in Wren AI or dbt) outperforms raw agentic reasoning by reducing response times from 20 seconds down to under 7 seconds… MotherDuck provides sub-second latency for analytical queries, positioning itself as the SQLite for Big Data for teams that find Datasette too limited for scale.

GitHub datasette/datasette-agent (auth and permissions) github.com

The agent inherits the permissions of the authenticated GitHub user. Executing write SQL or creating new tables requires explicit user approval and the execute-write-sql permission… 79% of developers grant AI agents access to cloud environments despite 97% expressing reservations about credential leaks.

blixt/phonon GitHub (independent WebGPU port) github.com

115x real-time on M3 Max, first audio in as little as 3 ms, with live word-level timings extracted directly from one of the model’s attention heads — something the official Gradium browser package does not expose.

Medium teardown by jinlow (‘The four things you hand over…’) jinlow.medium.com

The final synthesis happens locally, but the creation of a cloned voice file still requires Gradium’s cloud API — so Phonon is not a 100% offline solution for developers who need local-only cloning.

Hacker News discussion (item 50369215) news.ycombinator.com

Some users reported extreme distortion that sounded like ‘a warped cassette tape’ or ‘garbled words’, and the model consistently stumbles on rupee amounts, PIN codes, dates and acronyms like ‘GPUs’.

alphaXiv 2606.23335 — ‘Watermark Shortcut’ study alphaxiv.org

Detectors trained on watermarked-synthetic vs unwatermarked-real audio latch onto the watermark itself as the ‘fake’ label; adding an AudioSeal mark to a genuine recording pushed error rates from 16% to 75% (‘mark-to-frame’).

DataRoot Labs TTS comparison datarootlabs.com

Kokoro and Kitten TTS are StyleTTS2-based with GAN vocoders synthesising entire chunks; Kokoro’s first-audio latency is >500 ms versus Phonon’s ~7–8 ms frame-by-frame streaming — roughly a 60x gap on the same MacBook Air M5.

FermionResearch/Phonon-1 model card (HF) huggingface.co

Users must supply precomputed 512-dim speaker latents from the Gradium hub; the Transformers v5.21 implementation ships without a Mimi encoder, so zero-shot cloning from a raw recording is not possible out of the box.

Jack Sun

Jack Sun, writing.

Engineer · Bay Area

Hands-on with agentic AI all day — building frameworks, reading what industry ships, occasionally writing them down.

Digest
All · AI Tech · AI Research · AI News
Writing
Essays
Elsewhere
Subscribe
All · AI Tech · AI Research · AI News · Essays

© 2026 Wei (Jack) Sun · jacksunwei.me Built on Astro · hosted on Cloudflare