Paul Sawers

@jambo.bsky.social

• ex-TechCrunch. • {Forkable} - A newsletter all about open source: https://forkable.io/ • All pitches: forkable@pm.me • Anonymous tip-offs @ Signal: PSTC.08

Anthropic is testing a feature called “auto dream” for Claude Code, aimed at keeping its memory from filling up with stale or conflicting information. Some early users compare it to REM sleep, in terms of how it consolidates and cleans memory. 🔗 Full story here: tessl.co/fvd

Anthropic tests ‘auto dream’ to clean up Claude Code's memory

Anthropic is testing 'auto dream' for Claude Code to manage memory by reviewing and rewriting stored context, addressing issues of stale or conflicting information.

tessl.co

Anthropic is pushing Claude toward an “always-on” agent — it can run tasks in the background, use your computer, and be controlled remotely. It’s taken a cue from OpenClaw, for sure, but with tighter controls around access. 🔗➤Full story here: tessl.co/qfm

Anthropic makes Claude an ‘always-on’ agent, with lessons from OpenClaw

Anthropic is evolving Claude into an 'always-on' agent, with some lessons learned from OpenClaw along the way.

tessl.co

A new release from Cursor turns its internal security agents into reusable templates teams can deploy themselves. For developers, that could mean continuous security checks — but it also raises questions about how much trust to place in AI reviewing code. 🔗Read the story in full here: tessl.co/ie4

Cursor turns its internal security agents into reusable templates

Cursor offers reusable templates for security agents, enabling continuous codebase scanning and automatic vulnerability fixes, enhancing software security review processes.

tessl.co

AI coding agents can run commands and access files with the same permissions as the developer who launched them — raising the risk of unintended access to sensitive data. Safehouse is a new open-source macOS tool that runs agents in a sandbox to limit what they can reach. tessl.io/blog/safehou...

Safehouse sandboxes AI coding agents on macOS

Safehouse is an open-source tool that sandboxes AI coding agents on macOS, limiting their access to sensitive files and directories to enhance security.

tessl.io

Data from The Pragmatic Engineer suggests that developers prefer Claude Code, but Copilot dominates in large companies — a reminder that Microsoft's enterprise reach matters as much as developer sentiment. 🔗Read the story in full here: tessl.io/blog/develop...

Devs love Claude Code, but Microsoft’s Copilot has enterprise edge

While developers prefer Claude Code, Microsoft Copilot dominates in enterprises, highlighting a divide between individual preferences and corporate adoption.

tessl.io

So Cursor just rolled out “Automations,” which lets AI coding agents run in the background and react to things like PRs, Slack messages, or incidents. The idea is to hand off routine engineering tasks to always-on agents. 🔗Read the story in full here: tessl.io/blog/cursor-...

Cursor launches Automations for always-on coding agents

Cursor's Automations feature enables coding agents to run automatically in response to events, streamlining codebase maintenance and routine engineering tasks.

tessl.io

Anthropic has published its first Agentic Coding Trends Report, looking at how AI coding agents are reshaping software engineering. The core message is clear: agents handle more of the implementation work, while humans focus more on orchestration. 🔗Read the story in full here: tessl.co/sob

Anthropic: 8 agentic coding trends shaping software engineering in 2026

8 key agentic coding trends for 2026, according to Anthropic. Learn how AI agents are transforming the software development lifecycle.

tessl.co

Gabor Soter, founder at London GenAI consultancy Palindrom, says software engineering is changing so fast that waiting for best practices is riskier than moving fast — as long as teams adopt strong verification, security guardrails, and product judgment. 🔗 Full interview: tessl.io/blog/why-its...

Why ‘it’s better to overshoot than undershoot’ in the age of AI coding

An interview with Gabor Soter, founder of generative AI consultancy Palindrom, who plans to go all-in on agentic coding tools in 2026.

tessl.io

This is pretty cool — OpenAI has dropped the Codex desktop app, and it’s like a developer control panel for AI. You can have a bunch of coding agents working at the same time on different parts of a project, and let them handle long-running tasks without losing track. 🔗 tessl.io/blog/openai-...

OpenAI launches macOS Codex app for supervising AI agents

OpenAI launches new macOS Codex app for managing AI agents. Supervise parallel tasks, integrate with GitHub, and enhance your coding workflow.

tessl.io