Love the little request system here. Missing a figure? Ask on GitHub and the most asked for get drawn first. Plus now there's a skill to just draw your own.
https://t.co/nGC1hug8BG
J ai découvert un skill qui permet de créer des illustrations isométriques animées 👀
C est clairement ce genre d illustration qui fait super pro sur un site ! J ai pas eu le temps de tester le skill encore mais je veux bien voir vos résultats !
Le skill : https://t.co/TnKIpo08w8
Calling this a weekend list is wild lol, it's like two years of platform team roadmap. The $0.50 per query cost kill-switch is honestly the one to steal first though.
https://t.co/97qd7jzP8V
CANCEL your weekend plans.
You NEED to:
• Replace fragile chat loops with Temporal state machines for crash-resilient agent workflows
• Build a dynamic context assembler that strictly budgets tokens between memory, tools and docs per request
• Implement a System-One router to handle 95% of triage in milliseconds, escalating only edge cases to reasoning models
• Write a raw Model Context Protocol (MCP) server from scratch to expose your database securely to external agents
• Build trajectory grading in CI that blocks PRs if an agent's tool-call sequence deviates from the golden path
• Add an inter-agent sanitization proxy so Agent A's output can't execute a prompt injection on Agent B
• Implement a cost kill-switch that halts any agent loop if projected token spend exceeds $0.50 per query
• Build a 3-tier memory engine (Working, Episodic, Semantic) with automated eviction and compression policies
• Set up KV-cache prefix sharing at the proxy layer to slash Time-To-First-Token by 80% for identical system prompts
• Route 5% of production traffic to a new model silently, compare the trajectories and auto-generate a diff report
• Build an API-to-Browser fallback where the agent spawns a headless browser if the REST API 404s
• Automate a DPO pipeline: user thumbs-down → auto-format to preference pairs → queue a nightly LoRA fine-tune
• Implement a PII unmasking proxy that swaps sensitive data for UUIDs before the LLM sees it and restores it post-generation
• Set up a speculative decoding pipeline: local 2B model drafts tokens, cloud 70B verifies them, cutting latency 60%
• Build a 4-tier graceful degradation chain: Frontier API → Mid-tier → Local Quantized → Semantic Cache
• Add checkpointed human-in-the-loop approval gates wired directly into Slack for high-stakes agent actions
• Trace every LLM hop with OpenTelemetry-style spans capturing exact token counts, latency and tool arguments
• Fuzz your own multi-agent swarm with malformed tool outputs and context overflow to test self-healing recovery
You have way too much to do.
Bookmark & Repost.
Sergio Martinez has done a ton of art direction and graphics for Nike ACG and his powers are high as you imagine ACG looking. Outstanding work.
[Link↓]
Really like the idea that the chat is just an interface and the Task is what persists. Never having to dig for which conversation had the unfinished work would honestly be the whole appeal.
https://t.co/Hg7SlbQhUq
Introducing Comma.
For forty years, we have worked for our software. We learned its menus, gave it our time, and stayed at the keyboard until the work was done. Comma exists to turn that around.
No group chats. No crowd of assistants to babysit. Just Comma. Tell it what matters, and it carries the work through, across every device, until it's done.
Open source. Now live on macOS and the web, with more platforms coming soon.
https://t.co/6ufCWOWcCx
AI assistants as little glass keychain charms is so cute. And the note saying together they mostly add up to a lot of notifications lol
https://t.co/lcn2bg2S7C
These are so fun. Love the masked wrestlers just casually hanging out in living rooms and cafés, all drawn super loose with raw crayon color.
https://t.co/vNLmsvvPra
247Studio is a small polish studio working with the honesty of 2010. Maybe change the name, don't change the attitude. Everybody likes going back in time.
[Link↓]
Kind of love the idea of a music platform without the whole release framing. Most places make posting a rough demo feel like launching an album.
https://t.co/WmDkftDCrN
New from me; HEDROOM™ is a space for music makers to share their sound. I wanted to create a platform where people could share their work without the formalities and framing of traditional music platforms, and have a bit of fun while doing it. Now in public beta 🔗⬇️
Such a nice little recipe for fake print misregistration. Splitting the flower into separate color plates and nudging them apart gets really close to the real thing.
https://t.co/A9wnUcyeim