Daily Issue
Vol. I — No. 18
22 · 05
Friday, 22 May 2026
Generated 2026-05-22 12:13
google/gemini-2.5-flash-lite-preview-09-2025
飒爽英姿闯江湖，诗酒茶话莫孤独。 — 是二智呀 50 items · 4 sections
§ 0

The Morning

Local weather 1
This morning in
London
Clear sky
Today's range
27.9°↓16.1°
currently 26.6°
Feels
26.4°
Rain
0%
Wind
15 km/h
Humid
38%
Rise
04:58
Set
20:55
§ I

US Stocks

Pre-market signal radar 12
US pre-market radar
premarket 2026-05-22
0 Bullish
0 Bearish
12 Neutral
Sector Tape
Energy Infrastructure 1 names
64 Top: VST · Neutral · RS +4.0% Bullish 0 / Bearish 0 / 5d +5.1%
Servers and Thermal Management 2 names
58 Top: DELL · Neutral · RS -4.9% Bullish 0 / Bearish 0 / 5d -6.0%
Battery and Energy Storage 3 names
57 Top: EOSE · Neutral · RS -1.5% Bullish 0 / Bearish 0 / 5d -1.3%
Hyperscale Cloud 4 names
56 Top: ORCL · Neutral · RS -1.1% Bullish 0 / Bearish 0 / 5d -0.9%
Networking Equipment 4 names
55 Top: CIEN · Neutral · RS +0.8% Bullish 0 / Bearish 0 / 5d +0.3%
Foundry 2 names
54 Top: INTC · Neutral · RS +1.3% Bullish 0 / Bearish 0 / 5d -0.1%
Compute Mining 4 names
53 Top: HUT · Neutral · RS -3.3% Bullish 0 / Bearish 0 / 5d -3.2%
Manufacturing 4 names
52 Top: SANM · Neutral · RS -4.1% Bullish 0 / Bearish 0 / 5d -6.2%
Ticker Setup Move Score Evidence Quality
FLNC Fluence Energy Battery and Energy Storage
Neutral News watch Low confidence
+0.7% $20.33 5d -3.6%
58 sector flat RS -3.8%

Watchlist item from 2 recent headline(s).

Why Fluence Energy Stock Is Powering Higher - TipRanks Needs fresh price/news confirmation before becoming an actionable setup.
quote: delayed fallback news: fresh financials: fresh news: 2
quotes: nasdaq 24 24/24news: google_news_rss 23, gdelt 1 24/24filings: 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:2605.22763v1Lead article

Advancing Mathematics Research with AI-Driven Formal Proof Search

George Tsoukalas, Anton Kovsharov, Sergey Shirobokov, Anja Surina, Moritz Firsching

his paper introduces and evaluates a method where Large Language Models (LLMs) generate formal proofs in languages like Lean to overcome their inherent unreliability in mathematical reasoning. The core contribution is the first large-scale demonstration of this AI-driven formal proof search, showing agents autonomously solved 9 open Erdős problems and proved 44 OEIS conjectures, validating the approach for active mathematical research.

Read abstract →Full PDFLINNET · SCORE9RELEVANCE
Example inputs/outputs for an AlphaProof-equipped agent (applied to Erdős #125). The user provides a Lean file with a specification of the problem, and an empty proof body replaced with the sorry placeholder. (a) Modifications are permitted only within EVOLVE-BLOCK and EVOLVE-VALUE markers. (b) During sketch refinement, the prover subagent is shown an assembled prompt template with the current proof, and optionally prior attempts/sketches, their Elo ratings, and feedback from AlphaProof’s attempts on unsolved goals. (c) The prover reasons about the problem informally and invokes tools. In this example, the prover invoked AlphaProof which resolved all but one goal. The prover then decomposed that goal into three simpler lemmas, and called AlphaProof again, which then resolved all remaining goals. The agent also produced a natural language summary of its attempt at the end of generation.
Example inputs/outputs for an AlphaProof-equipped agent (applied to Erdős #125). The user provides a Lean file with a specification of the problem, and an empty proof body replaced with the sorry placeholder. (a) Modifications are permitted only within EVOLVE-BLOCK and EVOLVE-VAL…
Agentic CLEAR Pipeline. We start by preparing the execution traces. Stage 1: Apply multi-level per-trace evaluation via an LLM Judge. Stage 2: Aggregate insights using CLEAR, split into System-wide patterns and Node-specific patterns, and prepare them for the UI.
Agentic CLEAR Pipeline. We start by preparing the execution traces. Stage 1: Apply multi-level per-trace evaluation via an LLM Judge. Stage 2: Aggregate insights using CLEAR, split into System-wide pa…
cs.AIarxiv:2605.22608v1

Agentic CLEAR: Automating Multi-Level Evaluation of LLM Agents

Asaf Yehudai, Lilach Eden et al.

Agentic CLEAR is an automatic, dynamic evaluation framework designed to address the challenges of assessing complex LLM agent behavior. It provides multi-level textual insights into agent actions at the system, trace, and node levels, moving beyond basic obser…

abstractpdfLINNET · SCORE9RELEVANCE
cs.AIarxiv:2605.22714v1

AMEL: Accumulated Message Effects on LLM Judgments

Sid-ali Temkit

This paper introduces the "Accumulated Message Effect on LLM Judgments" (AMEL), demonstrating that the polarity of prior conversation history biases subsequent evaluations made by Large Language Models. Across numerous tests, models shifted their judgments tow…

abstractpdfLINNET · SCORE9RELEVANCE
Overview of AMEL. (a) Items where the model is uncertain at baseline absorb the most bias ( d = − 0.34 d=-0.34 ); confident-baseline items absorb less ( d = − 0.15 d=-0.15 ). (b) Negative context biases models more than positive context (paired per-item ratio 1.62 × 1.62\( \times \) , p < 10 − 39 p<10^{-39} ); marginal means yield ≈ 2 × \( \approx \) 2\( \times \) (Section 4.5 ). Even balanced history shifts models toward “no.” (c) Bias saturates immediately; 5 turns produce the same effect as 50.
Overview of AMEL. (a) Items where the model is uncertain at baseline absorb the most bias ( d = − 0.34 d=-0.34 ); confident-baseline items absorb less ( d = − 0.15 d=-0.15 ). (b) Negative context bias…
Mean conflict-insensitivity score (bars, left axis) and failure rate (line, right axis) by model. Based on 90 conversations per model.
Mean conflict-insensitivity score (bars, left axis) and failure rate (line, right axis) by model. Based on 90 conversations per model.
cs.AIarxiv:2605.22720v1

Can AI Make Conflicts Worse? An Alignment Failure in LLM Deployment Across Conflict Contexts

Andrii Kryshtal

This paper investigates the risk of Large Language Models (LLMs) exacerbating armed conflicts by generating harmful outputs like false equivalencies or genocide denial. The authors tested nine model configurations across 90 multi-turn conflict scenarios, findi…

abstractpdfLINNET · SCORE9RELEVANCE
cs.AIarxiv:2605.22662v1

Claw AI Lab: An Autonomous Multi-Agent Research Team

Fan Wu, Cheng Chen et al.

Claw AI Lab introduces an autonomous research platform that moves beyond single-agent pipelines by enabling users to instantiate and manage a customizable, multi-agent research team from a single prompt. Its core contribution is providing an interactive, labor…

abstractpdfLINNET · SCORE9RELEVANCE
Overview of Claw AI Lab. The system organizes automatic research into five connected layers: idea, planning, coding, experimentation, and writing layers. Each layer uses specialized agents and validation loops, while feedback can flow across layers to revise earlier decisions when needed.
Overview of Claw AI Lab. The system organizes automatic research into five connected layers: idea, planning, coding, experimentation, and writing layers. Each layer uses specialized agents and validat…
№06
cs.AI
9

Contractual Skills: A GovernSpec Design Framework for Enterprise AI Agents

Ting Liu

This paper introduces **Contractual Skills**, a design framework inspired by GovernSpec, to structure agent skills as inspectable, readable task contracts within enterprise AI syst…

abstractpdf
№07
cs.AI
9

DeltaBox: Scaling Stateful AI Agents with Millisecond-Level Sandbox Checkpoint/Rollback

Yunpeng Dong, Jingkai He et al.

DeltaBox addresses the bottleneck of slow state checkpoint/rollback (C/R) for stateful AI agents by proposing a change-based transactional C/R mechanism instead of full state dupli…

abstractpdf
№08
cs.AI
9

Post-Training is About States, Not Tokens: A State Distribution View of SFT, RL, and On-Policy Distillation

Dong Nie

This paper reframes post-training methods like SFT and RL not just by their loss functions, but by how they shape the **state distribution** used for learning. The core contributio…

abstractpdf
№09
cs.AI
9

Reducing Political Manipulation with Consistency Training

Long Phan, Devin Kim et al.

This paper addresses covert political bias in LLMs, where models handle opposing political topics asymmetrically. The authors introduce two metrics, Sentiment Consistency and Helpf…

abstractpdf
№10
cs.AI
9

Spreadsheet-RL: Advancing Large Language Model Agents on Realistic Spreadsheet Tasks via Reinforcement Learning

Banghao Chi, Yining Xie et al.

Spreadsheet-RL is a reinforcement learning fine-tuning framework designed to train specialized AI agents for complex, multi-step tasks within a realistic Microsoft Excel environmen…

abstractpdf
See all 20 papers →
§ III

The Town Square

Hacker News 9
790
pts
Top story
AI is just unauthorised plagiarism at a bigger scale

The article argues that current AI models fundamentally operate as large-scale, unauthorized plagiarism by remixing vast amounts of existing, uncredited training data.

axelk.ee21 May discuss on HN →
359
Shunning AI is the human choice
thehandbasket.co21 May
308
BBEdit 16
barebones.com21 May
297
News outlets are limiting the Internet Archive’s access to their journalism
niemanlab.org21 May
257
The memory shortage is causing a repricing of consumer electronics
davidoks.blog21 May
223
Samsung chip workers will get an average $340k bonus as AI profits soar
qz.com22 May
166
Steve Wozniak cheered after telling students they have AI – actual intelligence
businessinsider.com22 May
122
Multi-Stream LLMs: new paper on parallelizing/separating prompts, thinking, I/O
arxiv.org21 May
104
Tell HN: I'm tired of AI-generated answers
21 May
compiled overnight by google/gemini-2.5-flash-lite-preview-09-2025 · end of issue no. 18 · thank you for reading