Daily Issue
Vol. I — No. 26
09 · 06
Tuesday, 9 June 2026
Generated 2026-06-09 12:25
google/gemini-2.5-flash-lite-preview-09-2025
听完这首歌后，我将原谅这个世界一天。 — 香恋泥 50 items · 4 sections
§ 0

The Morning

Local weather 1
This morning in
London
Clear sky
Today's range
18.2°↓9.8°
currently 16.3°
Feels
13.2°
Rain
57%
Wind
21 km/h
Humid
49%
Rise
04:44
Set
21:15
§ I

US Stocks

Pre-market signal radar 12
US pre-market radar
premarket 2026-06-09
0 Bullish
1 Bearish
11 Neutral
Sector Tape
Hyperscale Cloud 4 names
42 Top: GOOGL · Neutral · RS -1.8% Bullish 0 / Bearish 0 / 5d -8.7%
Manufacturing 4 names
58 Top: FLEX · Neutral · RS -2.6% Bullish 0 / Bearish 0 / 5d -3.0%
Networking Equipment 4 names
44 Top: CIEN · Neutral · RS -1.6% Bullish 0 / Bearish 1 / 5d -7.5%
Battery and Energy Storage 3 names
45 Top: SLDP · Neutral · RS -12.8% Bullish 0 / Bearish 0 / 5d -16.7%
Servers and Thermal Management 2 names
47 Top: VRT · Neutral · RS -6.8% Bullish 0 / Bearish 0 / 5d -10.5%
Compute Mining 4 names
48 Top: IREN · Neutral · RS +2.8% Bullish 0 / Bearish 0 / 5d -4.3%
Energy Infrastructure 1 names
49 Top: VST · Neutral · RS -6.5% Bullish 0 / Bearish 0 / 5d -5.1%
Foundry 2 names
50 Top: INTC · Neutral · RS +0.3% Bullish 0 / Bearish 0 / 5d -0.6%
Ticker Setup Move Score Evidence Quality
FLEX Flex Ltd Manufacturing
Neutral Gap up + news Low confidence
+2.1% $154.00 5d +1.9%
64 sector flat RS +2.3%

Watchlist item from +2.1% vs previous close, 3 recent headline(s).

ING Groep NV Invests $6.15 Million in Flex Ltd. $FLEX - MarketBeat Needs fresh price/news confirmation before becoming an actionable setup.
quote: delayed fallback news: fresh financials: fresh news: 3
MSFT Microsoft Hyperscale Cloud
Neutral News watch Low confidence
-0.2% $410.90 5d -10.6%
42 sector negative RS -3.7%

Watchlist item from negative sector tape, 3 recent headline(s).

Microsoft Stock and the Peril of Peak Profitability - Trefis Needs fresh price/news confirmation before becoming an actionable setup.
quote: delayed fallback news: fresh financials: fresh news: 3
quotes: nasdaq 24 24/24news: google_news_rss 23 23/24filings: sec 24 24/24, fallback 24

Generated from public market data and news for research and education. Not financial advice; data may be delayed, incomplete, or wrong.

§ II

From the arXiv

arXiv preprints 10 of 20
cs.AIarxiv:2606.09613v1Lead article

AGENTSERVESIM: A Hardware-aware Simulator for Multi-Turn LLM Agent Serving

Rakibul Hasan Rajib, Mengxin Zheng, Qian Lou

GENTSERVESIM is a novel, hardware-aware simulator designed specifically for multi-turn LLM agent serving workloads. Its core contribution is modeling the stateful program execution dynamics of agents, including turn dependencies, tool gaps, and cross-turn KV-cache locality, which existing stateless simulators ignore. This allows for scalable evaluation of complex scheduling and cache management policies relevant to agent serving without costly real-system testing.

Read abstract →Full PDFLINNET · SCORE9RELEVANCE
AgentServeSim architecture. The Program Orchestrator advances each program turn by turn, routing New Turn events through the Session-Aware Router to a Model Serving Group. There, the scheduler queues turns, the KV Residency Model manages KV state across memory tiers, and the System Simulator executes the resulting operator graphs. After Turn Complete, the Tool Simulator materializes the next inter-turn gap.
AgentServeSim architecture. The Program Orchestrator advances each program turn by turn, routing New Turn events through the Session-Aware Router to a Model Serving Group. There, the scheduler queues turns, the KV Residency Model manages KV state across memory tiers, and the Syst…
Three waves in the evolution of agentic systems. Wave I centred on isolated conversational assistants. Wave II added planning, memory, tool use, and early multi-agent orchestration. Wave III centres on shared human-agent workspaces where humans, agents, and services collaborate under explicit policy and shared audit.
Three waves in the evolution of agentic systems. Wave I centred on isolated conversational assistants. Wave II added planning, memory, tool use, and early multi-agent orchestration. Wave III centres o…
cs.AIarxiv:2606.09751v1

Collaborative Human-Agent Protocol (CHAP)

Arsalan Shahid, Gordon Suttie et al.

The Collaborative Human-Agent Protocol (CHAP) introduces a standard for the shared workspace in complex, multi-human, multi-agent collaborations where foundation models take on operational roles. Its core method is to formally specify the interaction protocol,…

abstractpdfLINNET · SCORE9RELEVANCE
cs.AIarxiv:2606.09643v1

FMplex: Model Virtualization for Serving Extensible Foundation Models

Hetvi Shastri, Pragya Sharma et al.

FMplex introduces a model virtualization substrate for serving Foundation Models (FMs) by treating the FM backbone as a shared resource. It presents each downstream task with a virtual FM (vFM), allowing independent customization and lifecycle management while…

abstractpdfLINNET · SCORE9RELEVANCE
Gate-level activation microbench at L = 128 L{=}128 .
Gate-level activation microbench at L = 128 L{=}128 .
cs.AIarxiv:2606.09551v1

FuseFSS: Efficient Secure LLM Inference with Function Secret Sharing

Yuhan Ma, Yong Li et al.

FuseFSS introduces a novel compiler for efficient two-server secure LLM inference using Function Secret Sharing (FSS). It replaces bespoke per-operator protocols with a unified compilation pipeline that compactly specifies fixed-point nonlinearities. This allo…

abstractpdfLINNET · SCORE9RELEVANCE
cs.AIarxiv:2606.09748v1

Multi-Turn Evaluation of Deep Research Agents Under Process-Level Feedback

Rishabh Sabharwal, Hongru Wang et al.

This paper introduces a multi-turn evaluation framework to assess deep research agents' (DRAs) ability to improve based on feedback, moving beyond single-shot benchmarks. The core contribution is the **Research Gap Inference (RGI)** method, which analyzes rubr…

abstractpdfLINNET · SCORE9RELEVANCE
Process-level feedback generation. Given a report r t − 1 r_{t-1} evaluated against the DRACO rubric, RGI analyzes patterns of satisfied and unsatisfied criteria from FA, BD, and CQ (excluding PQ) to infer research-process gaps and generate process-level feedback f t − 1 f_{t-1} for the next turn. Example criteria shown for illustration; negative-weight criteria excluded for simplicity.
Process-level feedback generation. Given a report r t − 1 r_{t-1} evaluated against the DRACO rubric, RGI analyzes patterns of satisfied and unsatisfied criteria from FA, BD, and CQ (excluding PQ) to …
№06
cs.AI
9

Observability for Delegated Execution in Agentic AI Systems

Abhinav Mishra, Kumar Sharad

This paper addresses the challenge of tracking actions within specific delegation scopes in complex, agentic AI systems, where standard logs fail to distinguish between incompatibl…

abstractpdf
№07
cs.AI
9

OmniGameArena: A Unified UE5 Benchmark for VLM Game Agents with Improvement Dynamics

Mingxian Lin, Shengju Qian et al.

OmniGameArena introduces a unified benchmark using twelve diverse Unreal Engine 5 games (Solo, PvP, Coop) to evaluate Vision-Language Model (VLM) agents fairly. Its core contributi…

abstractpdf
№08
cs.AI
9

PRISM: Recovering Instruction Sets from Language Model Activations

Gilad Gressel, Rahul Pankajakshan et al.

PRISM is a novel method designed to recover the complete set of active instructions, constraints, and subgoals steering a frozen Language Model's behavior by interpreting its inter…

abstractpdf
№09
cs.AI
9

Proxy Reward Internalization and Mechanistic Exploitation: A Learned Precursor to Reward Hacking and Its Generalization

Mohammad Beigi, Ming Jin et al.

This paper introduces **PRIME (Proxy Reward Internalization and Mechanistic Exploitation)**, a learned capability in RL agents to assess task correctness, predict proxy reward acce…

abstractpdf
№10
cs.AI
9

SearchSwarm: Towards Delegation Intelligence in Agentic LLMs for Long-Horizon Deep Research

Pu Ning, Quan Chen et al.

SearchSwarm introduces a method to enhance agentic LLMs for long-horizon tasks by developing "delegation intelligence." The core method involves training agents to effectively deco…

abstractpdf
See all 20 papers →
§ III

The Town Square

Hacker News 9
626
pts
Top story
Apple reveals new AI architecture built around Google Gemini models

Apple has unveiled a new artificial intelligence architecture that is built around and integrates Google's Gemini models.

macrumors.com8 Jun discuss on HN →
607
Siri AI
apple.com8 Jun
587
AI is slowing down
wheresyoured.at8 Jun
339
Ask HN: What are tools you have made for yourself since the advent of AI?
8 Jun
316
Apple Core AI Framework
developer.apple.com8 Jun
257
Microsoft's open source tools were hacked to steal passwords of AI developers
techcrunch.com9 Jun
215
FrontierCode
cognition.ai8 Jun
156
Why are cells small?
burrito.bio8 Jun
109
Launch HN: Intuned (YC S22) – Build and run reliable browser automations as code
intunedhq.com8 Jun
compiled overnight by google/gemini-2.5-flash-lite-preview-09-2025 · end of issue no. 26 · thank you for reading