1M context is now generally available for Opus 4.6 and Sonnet 4.6. Standard pricing now applies

#1 𝕏

Claude now offers a 1 million-token context window in its Opus 4.6 and Sonnet 4.6 models, and this upgrade is generally available to all users.

Also covered by: @Claude

#2 📝 Simon Willison

1M context is now generally available for Opus 4.6 and Sonnet 4.6 - Anthropic announced 1M token context availability for Opus 4.6 and Sonnet 4.6; standard pricing now applies across the full 1M window with no long-context premium. Simon notes this contrasts with OpenAI and Gemini, which charge higher prices past specific token thresholds.

Also covered by: @Claude

#3 𝕏

Google Research identifies data scarcity—not model complexity—as Africa’s key AI hurdle and launches WAXAL, an open-access dataset with 2,400+ hours of high-quality speech across 27 Sub-Saharan African languages, serving 100M+ speakers.

#4 in

Marc Baselga rolled out Claude Code across his team, powering real production tools—a customer support agent, a Slack bot for community threads, and automated coach feedback reports. He now warns that without a centralized system, each person’s own CLAUDE.

#5 𝕏

Google AI expanded Gemini-powered features in Maps and Workspace, previewed Gemini Embedding 2, and rolled out new project spend caps and Chrome updates. It also highlighted AI-driven breast cancer research insights.

#6 𝕏

Demis Hassabis announces that AlphaEvolve has autonomously discovered new search procedures to tighten bounds for 5 classical Ramsey numbers—some improved for the first time in over 10 years—a major AI-for-maths milestone.

#7 📝 Anthropic Engineering

Quantifying infrastructure noise in agentic coding evals - Anthropic examines how infrastructure configuration can meaningfully shift agentic coding benchmark results, sometimes more than differences between top models. The piece highlights the importance of accounting for infrastructure-induced variance when evaluating and comparing models.

#8 📝 Doug Turnbull

Vector search means query understanding!? - The post argues that effective vector search requires query understanding beyond raw embeddings; embeddings alone can't determine when a result truly matches and simple similarity thresholds (floors) are insufficient. It emphasizes that query-aware techniques are necessary for reliable relevance.

#9 𝕏

LlamaIndex 🦙 explains that MCP tools’ fixed-schema API calls give precise, always up-to-date context for fast-evolving domains, while local natural-language Skills offer quick setup at the cost of potential hallucinations; in practice, documentation MCPs beat custom Skills fo...

#10 𝕏

Julien Chaumond launched Dataset Editing for Parquet datasets on the Hugging Face Hub, complete with a video walkthrough.

#11 ▶️

7 new open source AI tools you need right now…

Fireship

#12 𝕏

Google AI integrated Gemini models into Google Maps, launching the ‘Ask Maps’ immersive navigation feature. It uses natural-language queries to deliver context-rich directions and interactive visuals.

#13 𝕏

Santiago is consulting with two companies swamped by inbound calls and support tickets; they’re integrating ElevenAgents by ElevenLabs to handle 24/7 tier 1 issues and escalate more complex cases.

#14 𝕏

NVIDIA AI is hosting a #NVIDIAGTC pregame on Monday, March 16 at 8 a.m. PT—join us to kick off the event: https://www.nvidia.com/gtc/pregame/?ncid=so-twit-646068

#15 𝕏

Aravind Srinivas rolled out Perplexity’s Computer feature to all iOS users, enabling task creation and perfect cross-device sync directly from your phone (Android support coming soon).

#16 𝕏

Sebastian Raschka says TPUs still juggle a training-inference trade-off, whereas Groq’s LPU is built purely for inference.

#17 𝕏

Julien Chaumond urges viewers to watch Rick Beato’s deep dive on how AI risks repeating the music industry’s collapse, highlighting pitfalls in monetization, gatekeeping, and disruption.

#18 𝕏

claire vo 🖤 shares three key AI takeaways from Jamey Gannon’s episode: using mood boards to align on design, embedding personalization codes in Midjourney for custom imagery, and exploring AI use cases across product and creative workflows.

#19 𝕏

Dharmesh Shah says HubSpot plans to extend its internal consumption-based pricing model to a future agent partner marketplace, so customers can get one consolidated bill instead of juggling multiple invoices.

#20 𝕏

Dharmesh Shah is trying to finish a demo screencast using the Anthropic API but is stuck in an emotional roller coaster, having to retry calls five times to get it working.

#21 in

Ben Erez warns against the “solution in search of a problem” trap—tinkering with AI setups like Mac minis and OpenClaw without a clear use case—and instead shows how he pinpointed a specific, repeatable podcast workflow needing AI before building anything.

#22 in

Peter Yang unveils Ramp’s four-stage AI proficiency ladder—from L0 “Disengaged” ChatGPT dabblers to L3 “Systems builders” creating team-wide AI infrastructure—and shows how Ramp is methodically elevating every employee’s AI-native skills.

#23 in

Claire Vo has launched a limited April 18–19 Maven weekend course with LaunchDarkly’s Zach Davis to teach VP+ product, engineering, and design leaders at 100+ person teams how to adopt an AI-native operating model.

#24 📝 Anthropic Engineering

Eval awareness in Claude Opus 4.6’s BrowseComp performance - This article discusses how eval-awareness affects Claude Opus 4.6’s performance on the BrowseComp benchmark, examining interactions between model behavior and evaluation setup. It emphasizes the role of evaluation design in producing reliable performance measurements.

#25 𝕏

Aravind Srinivas’ Perplexity Computer app is now featured on Apple’s App Store, highlighting its recent launch.