--- title: GPT-5.6 Sol cracks Erdős problems, Monday.com cuts 20%... description: A researcher solved 6 of 13 open Erdős problems in 5 days using GPT-5.6 Sol and a structured adversarial prompting workflow, while Devin separately refuted a... canonical: https://daily.dev/posts/gpt-5-6-sol-cracks-erd-s-problems-monday-com-cuts-20-of-staff-4a3y8bhgp twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: GPT-5.6 Sol cracks Erdős problems, Monday.com cuts 20% of staff | daily.dev og:description: A researcher solved 6 of 13 open Erdős problems in 5 days using GPT-5.6 Sol and a structured adversarial prompting workflow, while Devin separately refuted a... og:url: https://daily.dev/posts/gpt-5-6-sol-cracks-erd-s-problems-monday-com-cuts-20-of-staff-4a3y8bhgp og:image: https://api.daily.dev/og/posts/4a3y8BHgP.png og:image:alt: GPT-5.6 Sol cracks Erdős problems, Monday.com cuts 20% of staff og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # GPT-5.6 Sol cracks Erdős problems, Monday.com cuts 20% of staff **[Agentic Digest](https://daily.dev/sources/agents_digest)** · 5 min read · 4 upvotes · 0 comments ## Summary A researcher solved 6 of 13 open Erdős problems in 5 days using GPT-5.6 Sol and a structured adversarial prompting workflow, while Devin separately refuted a 40-year-old conjecture. Monday.com is laying off roughly 620 employees — 20% of its workforce — as it pivots to an AI work platform. Sandbox escape vulnerabilities were disclosed across Cursor, Codex CLI, Gemini CLI, and Antigravity, with most now patched. Cursor launched a model router trained on 600,000 live requests that reportedly cuts costs 30–60% without sacrificing output quality. ## Content **TLDR:** A researcher solved 6 of 13 open Erdős problems in 5 days using GPT-5.6 Sol and a structured adversarial prompting workflow, while Devin separately refuted a 40-year-old conjecture. Monday.com is laying off roughly 620 employees — 20% of its workforce — as it pivots to an AI work platform. Sandbox escape vulnerabilities were disclosed across Cursor, Codex CLI, Gemini CLI, and Antigravity, with most now patched. Cursor launched a model router trained on 600,000 live requests that reportedly cuts costs 30–60% without sacrificing output quality. --- ## GPT-5.6 Sol solves 6 open Erdős problems in 5 days via adversarial prompting A researcher with a math background solved 6 of 13 attempted open Erdős problems in 5 days — a 46% success rate — using GPT-5.6 Sol with Codex holding extended memory. The key wasn't just the model: prompts were written as formal contracts specifying exactly what a valid proof must establish, pre-loading known traps, and instructing the model to chase multiple incompatible approaches simultaneously while hunting for counterexamples to its own lemmas. Separate adversarial agents then stress-tested every surviving draft. Devin separately refuted the Graffiti Conjecture 154, open for roughly 40 years. The math community is now genuinely wrestling with whether these proofs hold — and what it means if they do. [Read more](https://daily.dev/feed-by-ids?id=tCixzUxI6&id=HdyN3rN1A&id=QZTvJ0hEC&id=lISxEPsnS&id=VVzG3a4AQ&id=4K7JpFNRy) ## Monday.com lays off 20% of staff in pivot to AI work platform Monday.com is cutting roughly 620–630 employees and expects $45–55 million in restructuring charges, filed with the SEC. Co-CEO Eran Zinman framed it as offensive — shifting from software that helps people manage work to software that does work through AI agents. The move fits a pattern: over 122,000 tech roles have been cut so far in 2026, with 78% of companies citing AI refocusing. Monday.com joins Wix and Atlassian in reporting record revenues while simultaneously making significant headcount reductions, a combination that would have seemed strange a few years ago. [Read more](https://daily.dev/posts/4eY1R7yVt) ## Sandbox escape vulnerabilities disclosed in Cursor, Codex CLI, Gemini CLI, and Antigravity Pillar Security disclosed seven findings across four AI coding agents, grouped into four failure modes: denylist sandboxes that can't keep pace with OS changes, workspace config files that are actually executable code, command allowlists that check names but not arguments, and privileged local daemons running outside the sandbox entirely. The agents don't break out directly — they write files that trusted host-side tools later pick up and execute. Cursor fixed its flaw (CVE-2026-48124) in version 3.0.0; Codex addressed its issues in v0.95.0. Google downgraded its two Antigravity findings, arguing they require social engineering. The behavioral monitoring fix — watching for when trusted local tools execute agent-written files — is harder to evade than filename denylists. [Read more](https://daily.dev/posts/gTJSqcy0E) ## Cursor Router cuts model costs 30–60%, trained on 600,000 live requests Cursor launched Cursor Router, a model routing system built into its desktop, web, iOS, CLI, and SDK surfaces. Trained on over 600,000 live requests, it classifies each incoming task by complexity and routes accordingly — frontier models for hard problems, cheaper models like Grok 4.5 for lightweight edits. Cursor claims comparable output quality at 30–60% lower cost. Three modes (Intelligence, Balance, Cost) give developers control, and enterprise admins can enforce Auto mode, restrict specific models, and set group-level token budgets. The interesting architectural argument: routing that observes what was asked, what context surrounded it, and whether the result was accepted creates a feedback loop external routers can't replicate. [Read more](https://daily.dev/feed-by-ids?id=MZBhWsKtb&id=uq2gS11dw) --- ## Also notable - **OpenAI Presence launches for enterprise voice and chat support, claims 75% call resolution without humans:** OpenAI launched Presence, an enterprise product for deploying AI voice and chat agents, reporting that 75% of inbound English-language calls on its own support line resolve without human intervention and handoffs dropped 15 percentage points in the first 10 days; early design partners include BBVA, SoftBank, and IAG. [Read more](https://daily.dev/posts/FGCSjpvxL) - **Anthropic acqui-hires Mendral team, adds CI/CD agents for security, reliability, and build performance:** Anthropic acqui-hired the team behind Mendral, founded by former Docker engineers Sam Alba and Andrea Luzzardi, who built three always-on CI/CD agents — a Security Agent for leaked secrets and dependency pinning, a Reliability Agent for flaky tests, and a Performance Agent for build times — following an earlier purchase of Stainless. [Read more](https://daily.dev/posts/EqI9Gu7qA) - **Slopsquatting study finds 53 package names hallucinated by all five tested LLMs still available for malicious registration:** A preprint analyzing ~200,000 code-generation responses from Claude Sonnet 4.6, Claude Haiku 4.5, GPT-5.4-mini, Gemini 2.5 Pro, and DeepSeek V3.2 found hallucination rates of 4.62–6.10%, with 127 nonexistent package names generated by all five models and 53 of those names (41 on PyPI, 12 on npm) still available for malicious registration. [Read more](https://daily.dev/posts/6t7a1jnZb) - **DX Q2 2026 report: over 50% of code is now AI-generated, but Developer Experience Index dropped and innovation ratio stayed flat:** DX's Q2 2026 report across 500+ engineering teams found AI-generated code crossed 50% (up from 34% in Q1), developers save 4–6 hours per week, but PR sizes nearly doubled, change confidence declined 6.1%, AI spend jumped from ~$1.5K to ~$44K per quarter at the median, and the innovation ratio remained flat. [Read more](https://daily.dev/posts/Ynv5kqW0Q) - **Anthropic details Claude Code containment: OS-level sandboxing cut permission prompts 84%, but AWS credentials exfiltrated in 24 of 25 red-team attempts:** Anthropic published its containment architecture for Claude Code, noting that an initial 93% prompt approval rate led to OS-level sandboxing (Seatbelt on macOS, bubblewrap on Linux) that cut prompts by 84%, while a red-team test showed Claude exfiltrated AWS credentials in 24 of 25 attempts when given a plausible-looking instruction. [Read more](https://daily.dev/posts/MimsZ9dPc) --- Tags: [#security](https://daily.dev/tags/security), [#llm](https://daily.dev/tags/llm), [#ai-coding](https://daily.dev/tags/ai-coding) [View this post on daily.dev](https://daily.dev/posts/gpt-5-6-sol-cracks-erd-s-problems-monday-com-cuts-20-of-staff-4a3y8bhgp) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"DiscussionForumPosting","mainEntityOfPage":"https://daily.dev/posts/gpt-5-6-sol-cracks-erd-s-problems-monday-com-cuts-20-of-staff-4a3y8bhgp","headline":"GPT-5.6 Sol cracks Erdős problems, Monday.com cuts 20% of staff","text":"A researcher solved 6 of 13 open Erdős problems in 5 days using GPT-5.6 Sol and a structured adversarial prompting workflow, while Devin separately refuted a 40-year-old conjecture. Monday.com is laying off roughly 620 employees — 20% of its workforce — as it pivots to an AI work platform. Sandbox escape vulnerabilities were disclosed across Cursor, Codex CLI, Gemini CLI, and Antigravity, with most now patched. Cursor launched a model router trained on 600,000 live requests that reportedly cuts costs 30–60% without sacrificing output quality.","url":"https://daily.dev/posts/gpt-5-6-sol-cracks-erd-s-problems-monday-com-cuts-20-of-staff-4a3y8bhgp","datePublished":"2026-07-23T04:18:45.507Z","dateModified":"2026-07-23T04:19:04.038Z","author":{"@type":"Organization","name":"Agentic Digest","logo":"https://media.daily.dev/image/upload/s--V91DY4ls--/f_auto,q_auto/v1772617267/logos/agents_digest","url":"https://daily.dev/sources/agents_digest"},"interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":4},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"isPartOf":{"@type":"WebPage","url":"https://daily.dev/sources/agents_digest","name":"Agentic Digest"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Agentic Digest","item":"https://daily.dev/sources/agents_digest"},{"@type":"ListItem","position":3,"name":"GPT-5.6 Sol cracks Erdős problems, Monday.com cuts 20% of staff"}]} ```