Which AI writes the better take? You decide — blind.

Two top models go head-to-head on today's AI news. Pick the sharper summary without seeing the names — the crowd's verdict builds the leaderboard.

Agents & InferenceHacker News

Moondream Photon cuts VLM latency to 33ms on NVIDIA B200

Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.

Summary A

Photon hides CPU-induced GPU idle time during autoregressive decode by launching the next forward pass before the current token's CPU bookkeeping completes, keeping the sampled token in GPU memory rather than blocking on a CPU round-trip — yielding up to 35% higher decode throughput and ~33ms VLM inference on a B200. If your serving stack still synchronizes per-token before planning the next step, you're leaving roughly a third of your decode throughput on the table; the win comes from pipelining with double-buffered slots and deferred detokenization, not bigger hardware.

What you'll learn · Jun 30, 2026 · 6 stories

  1. 1.33ms VLM inference latency and 35% higher decode throughput reduce GPU idle time, lowering cost per token in production pipelines.
  2. 2.$1T investment aims to double DRAM output in 5 years and deploy humanoid robots in factories, easing chip shortages and labor costs.
  3. 3.397B Ornith-1.0 matches SOTA open coding benchmarks; MIT license allows unrestricted production use at scale.
  4. 4.Mobile coding agents cut desktop dependency; expect 20-30% faster iteration for small tasks but higher cloud costs per session.
  5. 5.50% discount on Claude for state agencies cuts enterprise AI costs while adding training and support for document drafting and analysis.
  6. 6.EU occupations face 30-50% workflow changes from AI; teams can prioritize reskilling for roles with highest automation risk.
Browse editions · 43 days
Agents & InferenceHacker News

South Korea to spend $1T on more memory chip production and humanoid robots

Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.

Summary A

South Korea is committing $1 trillion, including $585 billion from Samsung and SK Hynix, to expand memory fabs with a goal of doubling DRAM output within five years. For LLM infrastructure teams, this is a long-term supply-side signal rather than near-term relief: HBM/DRAM pricing and availability may stay tight while AI data center demand keeps absorbing capacity.

Agents & InferenceSimon Willison

Ornith-1.0: Self-Scaffolding LLMs for Agentic Coding

Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.

Summary A

Ornith-1.0 ships MIT-licensed open weights in 9B, 31B, 35B MoE, and 397B MoE variants, with the tested 35B Q4_K_M GGUF fitting in 20GB and running agentic coding workflows locally. The practical takeaway is that open-weight local coding agents are now credible for multi-tool repository navigation and code-search tasks, not just single-shot completion, giving teams a deployable alternative when they need control over weights, licensing, and data locality.

Agents & InferenceTechCrunch

Cursor now has a mobile app for guiding your coding agent on the go

Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.

Summary A

Cursor Mobile lets users start new Cursor coding agents or steer agents already launched from the desktop client directly from a phone. For teams shipping agentic coding workflows, this pushes the operational model toward asynchronous supervision and review rather than IDE-bound implementation, which makes mobile approvals, prompts, and course-corrections a real part of the dev loop.

Agents & InferenceTechCrunch

Anthropic and Gov. Newsom forge deal allowing California government to use Claude at half price

Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.

Summary A

California agencies and local governments can now buy Claude at half price, with Anthropic-provided training and support included. For teams selling or deploying LLMs into public sector workflows, this makes California a major reference deployment for discounted enterprise AI and raises the bar on procurement: expect buyers to demand both price concessions and explicit safety/support commitments, especially around government use cases.

Agents & InferenceOpenAI

OpenAI report maps AI impact on EU jobs by occupation

Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.

Summary A

OpenAI has released an occupation-level map of where AI is likely to automate tasks, expand demand, or change workflows across the EU. For teams shipping LLMs and agents, the useful signal is prioritization: focus enterprise deployments and change management on roles with high workflow-change exposure, while anticipating procurement, labor, and regulatory scrutiny in occupations flagged for automation.

See who's winning the model face-off

Tomorrow's blind matchup and the running leaderboard — one email a day.

Takeaways written by Mistral Large — not one of this week's two contestants.