← Dailies

Daily · 2026-11-02 00:00 UTC

DeepSeek ships a proactive coding agent; Adobe tests AI editing agents

DeepSeek open-sources a proactive coding agent, Adobe's EditAgent picks its own tools, National Grid secures land for Britain's biggest substation, and Singapore's artists push back on generative AI.

Source ↗

🎧 News Digest — 2026-11-02 00:00 UTC 9 items · ~2 min 50 s listen

1. DeepSeek open-sources a proactive coding agent — The dsh-superagent-prototype repository is a proactive coding agent for a single GitHub repository, packaged as a bundle for DeepSeek Harness. It listens to GitHub and optionally Feishu; when someone allowed asks for work, it opens a work item, works in its own agent session and directory, and answers in the original conversation — for example with a pull request. A web panel shows what needs a person. DeepSeek labels it a "research prototype for reference only, not an official DeepSeek product." GitHub · Nov 1
2. Adobe's EditAgent picks the right editing tool per step — Adobe Research and UC San Diego built EditAgent, which selects a specialised tool for each task instead of retrying one general-purpose editor, then lets a standard language model choose the best result from measured image data. The approach works, but needs about ten times the compute. The Decoder · Nov 1
3. Google's StreamVibe speaks up only when a task needs help — A Google-led research team built StreamVibe, a video agent designed to watch someone cook or assemble furniture and offer help only when it is needed. A benchmark found that none of the tested models reliably pick the right moment in longer videos, with errors compounding and response quality dropping sharply. The Decoder · Nov 1
4. National Grid buys Teesside land for Britain's biggest substation — National Grid agreed to purchase 29.5 hectares (73 acres) at Dorman Point on land owned by Teesworks, next to a planned AI datacentre, in a deal that could bring up to 3.2GW of capacity. Ben Houchen, mayor of the Tees Valley combined authority, said everyone who has tried to build a datacentre in the country knows the biggest problem is getting enough power to the site quickly. The Guardian · Nov 1
5. First Nations artist calls AI output "Temu slop" — Sally Scales, a Pitjantjatjara artist and chair of the National Gallery of Australia's First Nations advisory group, said AI is turning First Nations art into "Temu slop" and called it "a theft" of cultural knowledge that strips out lineage and lived experience. Searches on Temu, Shein and Amazon surface countless works described as Aboriginal paintings — plus coasters and stubby holders — mostly without attribution. The Guardian · Nov 1
6. Singapore's artists on protecting work from generative AI — CNA spoke with illustrators, animators, graphic designers, painters and arts educators about how they are responding to generative AI while preserving the human touch in their work. NTU associate professor Bernhard Schmitt noted animation has used machine learning and automation for decades; Kinetic Singapore founder Pann Lim described responses ranging from outright rejection to enthusiastic adoption. CNA · Nov 1
7. Flock keeps cameras up after cities cancel their contracts — Plymouth, Minnesota's city council voted to terminate its Flock Safety contract in September, received no acknowledgement for over a month, and its police chief directed staff to disconnect the cameras' power cords. Dayton, Minnesota was told removal could take up to 60 days, then found Flock had reconnected two cameras within city limits without authorisation and against a specific police request. The Verge · Nov 1
8. Prop 45 could hand data centres a fast track — The California ballot measure would set a 365-day deadline for environmental impact reports, require legal challenges to be settled within 270 days, allow only 45 days of public comment, and require consultation only with federally recognised tribes. Earthjustice senior attorney Fernando Gaytan warns data centre developers could "shoehorn their projects into the definition of an essential project" and skip meaningful scrutiny. The Verge · Nov 1
9. Security review is the bottleneck for agents in production — Protegrity's Diego Loureda said sensitive data placed in a model, a vector store or an agent's memory "is going to propagate to other agents," and that there is no practical way to remove it afterwards. His argument: tokenisation keeps protected data usable, and governance has to approve patterns rather than individual projects — 83% of leaders in a study Protegrity commissioned from Enterprise Management Associates reported project delays. SiliconANGLE · Nov 1

🧪 Try This Next

• 🔧 Clone deepseek-ai/dsh-superagent-prototype and read docs/000-philosophy.md — the "act only when asked, or when a team rule allows it" design is the reusable part.
• 🎥 Study how StreamVibe decides when to interrupt. Your agent has the same timing problem, and most get it wrong.
• 📝 Before the next agent deployment, write down which data classes must never enter a vector store or agent memory.

Content on Anagnorisis is summarized, paraphrased, and editorialized from publicly available sources for length and clarity. Original sources are linked where available. All trademarks belong to their respective owners.