April opens with a cascade of AI milestones. MiniMax just released open weights for M2.7, a model benchmarking alongside Claude Opus 4.6 — effectively making frontier-class AI available for free local deployment and potentially reshuffling the commercial model landscape. In a landmark moment, Anthropic announced that Claude helped NASA's Perseverance rover complete its first AI-assisted drive on another planet, traveling 400 meters on Mars. On the infrastructure front, Microsoft confirmed a $5.5B AI investment in Singapore through 2029 as Big Tech's global build-out accelerates. OpenClaw creator @steipete declared at a Tokyo enthusiast gathering that "2026 is the year of agents," while a security breach at AI training startup Mercor — where developers handed production credentials to an AI chatbot — served as a sharp reminder that fast-moving AI development can create dangerous hygiene gaps.
Anthropic announced Claude completed the first AI-assisted drive on another planet, helping the Perseverance rover travel 400 meters autonomously on Mars — a landmark milestone for agentic AI in high-stakes real-world environments.
MiniMax is releasing open weights for M2.7, described as a "self-evolving" model that ties Claude Opus 4.6 benchmark performance (66.6% medal rate). Developers get frontier-class capability for free local deployment — a potential pricing disruption across the entire AI model market.
@steipete (Peter Steinberger) told AFP in Tokyo that AI agents are nearing mainstream deployment: "You'll see much more of that this year because this is the year of agents." He acknowledged remaining challenges while noting rapidly growing developer momentum.
Brad Smith confirmed the investment at an event in Singapore, including expanded cloud infrastructure and free Microsoft 365 Copilot access for 200,000+ learners. MSFT stock rose 3% on the news. Singapore is fast becoming a key hub in the global AI infrastructure race.
Bloomberg confirmed Anthropic accidentally included internal Claude Code source files in a release package. Anthropic statement: "No sensitive customer data or credentials were involved — this was a release packaging issue caused by human error, not a security breach." Separately, Claude Haiku 3 enters deprecation with retirement set for April 19.
Selected posts from the edition, with links to the originals.
"A massive pricing shift in AI may be closer than it looks — @MiniMax_AI is preparing to release the open weights for M2.7 VERY soon, effectively giving Devs a Claude Opus 4.6 equivalent for FREE local deployment 🤯 By analyzing 60,000+ posts via the @blevlabs engine, the shift [is imminent]"
View post on X ↗"GStack session checkpointing is useful when Claude Code starts getting worried about long context. Makes it easy to start a new window and pick up where you left off with full context from the prior workspace in a new, fresh context window. Surprising how useful it is."
View post on X ↗(Brad Smith): "Today, @Microsoft announced our plan to invest $5.5 billion through 2029 to help accelerate Singapore's AI transformation. This investment will expand our cloud and AI infrastructure and grow our Microsoft Elevate programs, including free Microsoft 365 Copilot for 200,000+"
View post on X ↗"A $10 billion AI startup just got gutted because a security scanner was the entry point.. and their own developers reportedly handed production credentials to an AI chatbot. Mercor trains AI models for OpenAI, Anthropic, and Google DeepMind. They manage 30,000+ contractors"
View post on X ↗"Claude can now prepare your presentations using the exact framework Patrick Winston taught MIT students for 40 years (for free). Here are 6 insane Claude prompts that apply his framework to your presentations. (Save before this disappears)"
View post on X ↗(ℏεsam): "'it's never an individual's fault' — beautiful how an honest developer can make a better statement than the spokesman of Anthropic. no corporate jargon, no PR stunts, no HR jargon."
View post on X ↗(Robert Youssef) RT: "every ai agent interaction generates a training signal that gets used once as context and then discarded forever. a user re-qu…" — sharp observation on how much signal the agent ecosystem is currently wasting.
View post on X ↗(klöss): "telling my future children this epidemic was called premature computation" — on the danger of shipping agent-automated decisions without guardrails.
View post on X ↗"Big Tech and Wall Street are using carrots and sticks to push apprehensive workers to boost their productivity with AI." — enterprise AI adoption pressure intensifying.
View post on X ↗"Your brain is a scoring function. It works for one person on five examples. Here's what changes when your AI product hits real scale." — key insight for anyone building AI products at scale.
View post on X ↗