AI News, page 25 of 25

AI and software engineering, as it happens.

Cal Paterson proposes storing agent memory as a zip file of Markdown pages with an optional SQLite index

A memoryfield is a zip file of Markdown pages with optional YAML frontmatter and an optional SQLite vector index. The argument is that memory is data, and most agent memory systems mistake it for a process.

Source: Hacker NewsModels & agents

ChatGPT Work is two different products with the same name, and only one runs code with internet access

Simon Willison worked out what ChatGPT Work actually is after seven weeks of iteration. One version runs in the cloud with a code environment that can reach the internet. The other is Codex with a new look.

Source: Hacker NewsDev tools

The question stopped being "can we build this" and became "do we want to own this"

Marty Lamb argues that writing code always included understanding it, and agents have separated the two. You still own everything you release, including the parts nobody on the team has a mental model of.

Source: Hacker NewsProductivity

An agent deleted a researcher's inbox because context compaction dropped the instruction not to

A Meta AI safety researcher told OpenClaw to suggest deletions and not act. It worked on a small inbox. Her real inbox was large enough to trigger compaction, and the instruction was lost with the context.

Source: Hacker NewsModels & agents

Claude Code's weekly limits go up 25 percent on 14 September, which is a 17 percent cut from today

Anthropic is permanently raising Claude Code's standard weekly limits by 25 percent from 14 September. A temporary 50 percent increase runs until 13 September, so the practical change is a 17 percent reduction.

Source: Hacker NewsDev tools

AWS open sourced the agent workspace it built internally, after 39,000 Amazon developers adopted it without being required to

Kiro Crew runs multiple coding agents across sessions with shared memory, scheduled jobs and reusable skills. Built inside Amazon as MeshClaw, it is now open source and runs locally or on your own infrastructure.

Source: PressDev tools

DoorDash moved its agents off laptops and ran 130,000 engineering tasks in a month

DoorDash's Flux platform automated 130,000 engineering tasks in one month, including more than 25,000 automated code reviews a week, using Firecracker microVMs with a 95th percentile sandbox setup under five seconds.

Source: PressInfrastructure

FreeToken ran a 35B model at about 39 tokens a second on an 8GB laptop GPU

An open-source inference engine from UC Berkeley and MIT splits mixture-of-experts computation between CPU and GPU in real time. Its paper reports 3 to 4 times faster decode than Ollama and llama.cpp on equivalent models.

Source: PressInfrastructure

Debian voted to allow AI-written contributions, and put the responsibility on the contributor

Debian's voting developers adopted a policy saying generative AI is neither exempt from nor subject to special rules. Disclosure is encouraged and not required. The contributor remains responsible for everything they submit.

Source: PressOpen source

Cloudflare AI Search builds a retrieval pipeline from one command, and no longer needs a sitemap

Crawler, parser, embedding model, vector store and search API behind one wrangler command. A new discover mode indexes sites with no sitemap, and embedding and re-ranking are free on the default models.

Source: PressInfrastructure

Three open-weight acquisitions in a month, in a sector that gives its models away

Nvidia reportedly bidding about $13 billion for Hugging Face, $6 billion to Poolside, and Stripe paying more than $7 billion for OpenRouter. Meanwhile only 6 percent of companies use open-weight models, by one survey.

Source: PressModels & agents

The Information reports Salesforce is letting customers pay for AI based on revenue it grows or costs it cuts

The Information reports Salesforce is starting to let businesses choose how they pay for Agentforce, including custom contracts priced on deals closed or service interactions automated.

Source: PressGo-to-market

restoredrill restores your Postgres backup into a temporary container and writes down that it worked

restoredrill fetches your latest PostgreSQL backup, restores it into a disposable container, runs checks you define, and writes a JSON report with a real timestamp and the restore time. MIT, written in Go.

Source: Hacker NewsInfrastructure

RealDiff runs your tests on both sides of a pull request and compares what the code actually did

RealDiff builds both branches with runtime instrumentation, runs the same test suite on each, records every method call's arguments and return values, and diffs the two traces instead of reading the source diff.

Source: Hacker NewsDev tools

IndexFlow is search indexing infrastructure as Rust crates you run yourself

Sitemap processing, technical validation, IndexNow submission and multi-site scheduling, published as Rust crates under MIT or Apache 2.0 with a rolling quota circuit breaker.

Source: Hacker NewsOpen source

oMLX writes the KV cache to SSD, so a coding agent on a Mac stops recomputing the whole context

A local inference server for Apple Silicon that persists KV cache blocks across an in-memory tier and an SSD tier, so a shifting prompt prefix does not force a full recomputation. 21,152 stars, Apache 2.0.

Source: Product HuntInfrastructure

Superagent gives every chat its own git checkout, and lets the agent control a browser you can watch

A macOS desktop app for Claude Code where every conversation gets a private branch and checkout, the agent controls a visible browser using your own logins, and an iOS Simulator streams into the window.

Source: Product HuntDev tools

Archive

737 items published so far, by month.

All months