cuppa

brewed 03:28 AM
← May 24

Monday

may25

2026

May 26 →

the brief

Agent and infra tooling took the spotlight: Emulate 0.6 ships deep Slack emulation for CI and Next.js agent sandboxes, Datasette 1.0a30 adds an extensible Jump menu, and Hyper debuts an API framework “as source” with MCP support. Local model performance jumps with llama.cpp’s MTP, while Anthropic tests file-based memory. Fresh datasets and evals landed, plus a hiring pulse on forward deployed engineering.

the poursit · sip · 12 items

pulse

(05)
  • @testingcatalogMay 24, 04:01 PM

    Claude prepping file-based memory

    Anthropic appears to be adding a Memory Files option for Claude—editable, organized notes it writes and recalls when relevant—promising more controllable, durable agent context.

    ANTHROPIC 🔥: Claude will soon receive a new file-based memory upgrade, offering users the option to choose between Memory Files and Classic memory. > Organized notes Claude writes as you chat and reads when they're relevant. Browse and edit them anytime. This feature appears pic.x.com/lEfGq0xFm8

  • @ctatedevMay 24, 05:55 PM

    Emulate v0.6 adds deep Slack support

    New release emulates Slack in CI and agent sandboxes with a Next.js adapter, covering messages, threads, channels, OAuth, events, files, and more for robust agent integration testing.

    New in emulate v0.6: way more Slack Emulate Slack in CI + agent sandboxes CLI or Next.js adapter Messages Threads Channels DMs MPIMs OAuth Inspector SDK tests Scopes Profiles Presence Files Uploads Pins Bookmarks Apps Modals Webhooks Events pic.x.com/dhqdOhLR1w

    signal 7hype 2sdk_updateslack_emulationagent_testingsource ↗
  • simonw/blog· feedMay 24, 11:52 PM

    Datasette 1.0a30 ships extensible Jump menu

    The new customizable Jump to… menu (triggered by “/”) improves navigation and plugin UX, powered by a JavaScript plugin hook for tailored shortcuts across Datasette instances.

    datasette 1.0a30 — <p><strong>Release:</strong> <a href="https://github.com/simonw/datasette/releases/tag/1.0a30">datasette 1.0a30</a></p> <p>The big new feature in this alpha is a new customizable "Jump to..." menu, described in detail in <a href="https://datasette.io/blog/2026/jump-menu/">The extensible "Jump to" menu in Datasette 1.0a30</a> on the Datasette blog. You can try it out by hitting <code>/</code> on <a href="https://latest.datasette.io/">latest.datasette.io</a> - it looks like t...

    signal 7hype 1release_notesalpha_releaseopen_sourcesource ↗
  • @unknownMay 24, 01:04 PM

    Hyper launches API framework as source

    Bun-based toolkit generates a runtime, OpenAPI, typed client, and MCP integration from your routes—no framework dependency—letting teams incrementally add auth and rate limiting as code.

    Hyper - an API framework as source, not a dependency ⚡ Built on Bun. Inspired by @shadcn - Your code, your repo. No framework in package.json - One route → runtime + OpenAPI + typed client + MCP - Add only what you need: `hyper add core auth-jwt rate-limit` bun create hyper pic.x.com/kFMycwdEBh

  • @tom_doerrMay 25, 01:04 AM

    Natural language simulates iOS and macOS users

    A new tool claims to execute plain-language goals on Apple platforms, useful for agent testing and end‑to‑end automation across desktop and mobile workflows.

    Simulates iOS and macOS users from plain language goals github.com/awizemann/harn… pic.x.com/GPkOMcD1xo

findings

(04)
  • @unknownMay 24, 10:12 PM

    llama.cpp MTP accelerates local inference

    Multi-token prediction support nearly doubles throughput on Qwen3.6‑27B (25→45 tok/s on A10G), pushing larger local models toward practical, daily-driver performance.

    llama.cpp with MTP support makes local models fast enough to use as daily drivers 🚀 Qwen3.6-27B dense generation below on A10G: From 25 tok/st to 45 tok/s (+78%)! pic.x.com/rLjBVa3Yzh

    signal 7hype 3llama_cppmtpperformance_benchmarksource ↗
  • hn/frontpage· feedMay 24, 04:31 PM

    Memory now dominates AI chip costs

    Epoch analysis finds memory accounts for nearly two‑thirds of AI chip component cost, underscoring HBM capacity and bandwidth as the key constraints for model scaling.

    Memory has grown to nearly two-thirds of AI chip component costs — Article URL: https://epoch.ai/data-insights/ai-chip-component-cost-shares Comments URL: https://news.ycombinator.com/item?id=48258684 Points: 312 # Comments: 335

    signal 6hype 1hardwarememorycost_analysissource ↗
  • @HuggingModelsMay 25, 02:20 AM

    Benchmark probes position bias in context

    A lightweight suite based on “Lost in the Middle” tests how model accuracy shifts as relevant facts move within the prompt, enabling targeted evaluation of position sensitivity.

    Based on the 'Lost in the Middle' paper (arxiv 2307.03172), this suite focuses on position bias. It's a lightweight, targeted benchmark for testing how model performance changes as relevant info moves from start to middle of input.

    signal 6hype 1benchmarkevaluationposition_biassource ↗
  • @unknownMay 24, 04:56 PM

    300k‑builder hardware survey released

    Hugging Face shares aggregated hardware profiles from 300,000 AI builders, offering a timely snapshot of GPUs, CPUs, and local setups powering the surge in on‑device AI.

    300,000 AI builders filled their hardware profile on @huggingface and we're sharing the results: https://t.co/3rLqeJGUCO. Excited to see how it evolves in the coming months especially with the explosion of local AI! pic.x.com/DB7Pk9jmLI

voices

(03)
  • pragmatic/engineer· feedMay 24, 08:35 PM

    Forward deployed engineering demand is back

    The Pragmatic Engineer reports FDE hiring heating up at Google, OpenAI, and Anthropic, signaling renewed emphasis on hands‑on, customer‑proximate engineering for AI deployments.

    The Pulse: Forward deployed engineering heats up again — Massive demand for the role at Google, OpenAI, and Anthropic. The latest version of the FDE role looks like the consultant / solution architect role done by many early-junior engineers.

    signal 6hype 2hiring_trendsfde_roleindustry_pulsesource ↗
  • @mattpocockukMay 24, 07:56 PM

    Using /handoff to span repositories

    Matt Pocock shows how Claude’s /handoff can move tasks across repos, a practical workaround when monorepo consolidation isn’t feasible and OSS fixes need coordinated follow‑through.

    Badass use case for /handoff I just discovered: Use it to break out of project directories and handoff tasks to other repos Great for: - "this really should be a monorepo but isn't" - "fix this bug in this OSS library"

    signal 4hype 3workflow_tipagent_handoffmulti_reposource ↗
  • @trq212May 24, 03:49 PM

    Legacy codebases as distillation fuel

    A takeaway from the Bun rewrite argues old code provides rich material for translating and modernizing software across platforms, potentially reducing reliance on legacy stacks like COBOL.

    my main takeaway from the Bun rewrite is that legacy codebases will be incredibly valuable as a source for "distilling" code into new forms every game should be crossplatform, all legacy software should work on the web, we don't need COBOL anymore

    signal 4hype 3opinioncode_distillationlegacy_softwaresource ↗