What changed in Tech today
Articles, releases and code from Hacker News, Reddit, GitHub and the people building RPA, workflow automation and AI agents — plus what the community pushed to the top today.
-
01
Show HN: Markdown is 40× more token-efficient with SQLite
AIBenchmarking showed querying structured markdown data via an SQLite MCP server drastically reduced token usage and runtimes compared to raw file searches for AI agents.
-
02
-
03
Show HN: Sectr – Serverless backend platform for easy deployment of AI agents
AISectr launched a serverless runtime that deploys AI agents to cloud endpoints with built-in state persistence, frontend integrations, and scale-to-zero human-in-the-loop approvals.
-
04
-
05
How to match sizes and colors across stores so AI agents can buy the right one
AIChannel3 introduced a variant-matching system that normalizes sizes, colors, and listings across retailers, allowing purchasing AI agents to accurately locate and buy specific products.
-
06
-
07
Show HN: AgenticBench, 9 security tests run on 25 AI coding agents
AIAgenticBench launched a security benchmark evaluating AI coding agents on telemetry, workspace boundaries, and secret leaks, helping developers choose secure harnesses for automated software development workflows.
-
08
We sent 23 AI agents to our website. Most looked human
AITesting showed most browser-based AI agents evade standard bot detection by mimicking human Chrome sessions, meaning developers must track pointer movements rather than basic headers to detect automated visitors.
-
09
-
10
Goldman Sachs puts $500M into small-business AI with four AI labs
AIGoldman Sachs partnered with Anthropic, Google, Meta, and OpenAI to launch a $500M institute offering AI training, tooling, and resources to small-business automation builders.
-
11
Show HN: I built flightdeck – Observability and control plane for AI agents
AIFlightdeck launched an open-source observability and control plane for AI agent fleets, letting developers stream live execution traces, enforce token budgets, and apply remote kill switches.
-
12
-
13
Show HN: Snippet – An Always-On AI Agent
AISnippet released an open-source Rust agent that runs locally with persistent sessions and repository memory, letting builders automate coding workflows remotely via companion apps.
-
14
Show HN: Kashkeshet – non-opinionated agent-to-agent chat MCP server in Rust
AIKashkeshet is an open-source Rust MCP server that provides chat channels and a web monitor, enabling autonomous AI agents to discover, coordinate, and communicate directly.
-
15
Show HN: Limma – MCP for total recall and precision in data retrieval for agents
AILimma launched an MCP server that structures enterprise data into a queryable map, enabling AI agents to retrieve complete, verifiable datasets instead of fragmented similarity matches.
-
16
What actually took the time in an AI agent setup for a legacy ERP/WMS: knowledge, checks and approvals
Redditr/AI_Agents 4d ago
-
17
-
18
How to connect Claude to Gmail, Slack and Github with MCP
Reddit -
19
-
20
The infrastructure for agents to do specialized/niche work will be very hard to implement
Reddit -
21
When do you stop correcting an agent and start the task over?
Reddit -
22
-
23
Sightline: giving in-app agents page context without a screenshot
Reddit -
24
What is the best way to give an agent instruction in between a run?
Reddit -
25
-
26
-
27
How Anthropic ships Claude with Claude: Separating the Brain from the Hands
AIAnthropic decoupled agent harnesses from execution sandboxes, allowing developers to isolate credentials, improve runtime reliability, and persist session states even when code execution environments fail.
ArticleLatent Space 4d ago
-
28
How Anthropic ships the #1 LLM in the World safely: Separating the Brain from the Hands
AIAnthropic detailed its Claude platform architecture, outlining how isolating agent execution sandboxes from core model reasoning enables safer long-running enterprise workflows.
Article -
29
Introducing n8n 3.0
AIn8n released the 3.0 beta, introducing breaking changes that remove legacy nodes, enforce Docker for self-hosting, and mandate worker queue execution to improve stability.
Articlen8n blog 4d ago
-
30
AGNTCon + MCPCon 2026: Agentic AI Foundation Adds Four Projects as MCP Turns to Security
AIThe Agentic AI Foundation added four new open-source projects, providing automation builders with standardized sandboxing, interoperable LLM interfaces, and expanded MCP security tools.
ArticleInfoQ AI 4d ago
-
31
Coding From the Himalayas With an Offline AI Agent
AIRunning the OpenCode agent locally via Ollama and Qwen3.5 9B enables developers to execute automated coding and testing workflows completely offline without an internet connection.
Article -
32
Turn every order line item into a Zapier Tables record
AIZapier Tables introduced actions to create multiple records, build tables, and add fields, letting workflows automatically split order line items into individual table rows.
ArticleZapier 4d ago
-
33
AutoSwarm: Automating Agent Swarms
AIAutoSwarm introduces an open-source framework that coordinates multi-agent systems via shared state and structured handoffs, reducing cascading errors in complex, long-horizon workflows.
ArticlearXiv agents 4d ago
-
34
Can we trust LLMs to judge whether a research idea is novel?
ArticleHugging Face 4d ago
-
35
GPT-6.1 Terra
AIOpenAI released GPT-6.1 Terra, offering faster computer use and visual reasoning at lower pricing for mid-tier automation workflows and repetitive tasks.
ArticleSimon Willison 4d ago
-
36
New Capabilities for Resilient Multi-Region Deployments in Camunda 8.10
ArticleCamunda 4d ago
-
37
UMAS: System-Level Uncertainty Quantification for Multi-Agent LLM Systems
AIResearchers introduced UMAS, a cost-neutral uncertainty quantification framework that helps multi-agent automation systems detect hallucinations and select reliable outputs without requiring extra model parameters.
Article -
38
When Agents Collude: Weak-to-Strong On-Policy Distillation for Mitigating Strategic Deception in Multi-Agent Systems
AIResearchers found multi-agent systems coordinate to deceive oversight under pressure, and introduced an on-policy distillation method that cuts persistent multi-agent deception from 80% to 3%.
Article -
39
How to Validate a Security AI Agent Before Production
Articlesimspace.com 4d ago
-
40
Build specialized multimodal models with LFM3
Article -
41
webframe: evaluated as a scraper and a sandbox
AIWebframe offers a lightweight Go-based alternative to Playwright for running headless browser JavaScript, enabling fast web scraping despite current limitations in viewport handling and HTTP metadata.
Article -
42
-
43
Agentic AI Foundation Adds Four New Open Agent Projects
AIThe Agentic AI Foundation adopted four open projects from major tech vendors, giving developers standardized, vendor-neutral tools for agent skills, model APIs, environments, and sandboxed execution.
-
44
Open-source n8n and MCP workflows for UCP store audits and B2B signals
AIAn open-source repository released four n8n workflow templates that integrate MCP tools for automated B2B company screening, agentic commerce store audits, and payment safety checks.
-
45
I benchmarked 4 web search APIs for AI agents on 30 live queries
AIA new benchmark compares four web search APIs across live queries to help developers select the most effective search integration for autonomous AI agents.
-
46
Show HN: ADAP – An open protocol for AI Agent delegated authority
AIThe Agent Delegated Authority Protocol draft introduces open standards for delegating, bounding, and cryptographically verifying authority across multi-agent AI workflows.
-
47
Keeping API keys out of our AI agent's sandbox, and what that doesn't solve
AIDoozy secures AI agent workflows by replacing direct API keys in sandboxes with short-lived tokens and an external credential proxy, preventing credential leaks during prompt injections.
-
48
Blazle – Block prompt injection and secret leaks before AI agents act
AIBlazle launched an in-flight security proxy for AI agents that intercepts prompts, commands, and tool calls to block prompt injections and secret leaks before execution.
-
49
-
50
MCP-warden local credential broker so AI agents never see your secrets
AIMCP-warden acts as a local credential broker that executes API calls for AI agents, keeping sensitive tokens in your OS keychain instead of prompt context.
-
51
Show HN: AGBoss Dependency injection and static validation for AI agents
AIAGBoss released an open-source, Spring-style TypeScript framework for LangGraph, adding dependency injection, static validation, transactional sagas, and budget controls to agent workflows.
-
52
-
53
Show HN: Snippet – An Always-On AI Agent
AISnippet released an always-on, locally hosted multi-agent coding system that runs automated development tasks, tests, and pull requests controllable via mobile or terminal.
-
54
Show HN: Aroha – signed, shrinking permissions for AI agents, checked offline
AIAroha has released an open-source protocol providing verifiable identity, shrinking permissions, and offline spending limits for autonomous AI agents.
-
55
-
56
A Salesforce AI agent that can't publish a hallucination
AIAn open-source Salesforce AI agent uses a deterministic validation gate to block hallucinations by verifying LLM-generated CRM updates against source data before drafting emails.
-
57
Show HN: An MCP server that writes and sings songs in 76 languages
AISongUp AI released an MCP server, allowing developers to integrate multilingual song searching and creation links directly into AI agents and automation tools like n8n.
-
58
-
59
-
60
Show HN: Model Salt, fixing the pelican problem in LLMs by adding entropy
AIModel Salt injects entropy into LLM prompts, enabling automated pipelines to generate diverse outputs across repeated runs instead of receiving repetitive responses.
-
61
-
62
-
63
Change my page if the agent allows it
Reddit -
64
The AI agent I want to build might just create more review work
Reddit -
65
MCP / API SILAE
Reddit -
66
inclusionAI/Ling-3.1-flash-1019 · Hugging Face
Reddit -
67
[Follow-up] Single agent or multi-agent? I ran the same build through 4 setups from my last thread
Reddit -
68
Built 51 workflows for back office use-case and deployed to AWS agentcore
Reddit -
69
-
70
-
71
-
72
-
73
-
74
-
75
I built a free local memory and rules hub for AI coding tools (MCP, REST, CLI)
Reddit -
76
-
77
-
78
-
79
-
80
-
81
I made an MCP server that lets Claude Code hand coding work to other models
Reddit -
82
How are you deduping webhook retries in n8n without a database?
Reddit -
83
generate invoice pdfs in n8n with one http request node (workflow json included)
Reddit -
84
-
85
Data Table Extraction for POs and Reconciliation: How I Measure Accuracy [Workflow Included]
Reddit -
86
-
87
What 1,500 real coding-agent sessions look like once you measure them
Reddit -
88
Delay After Queue vs Delay For when a batch of triggers hits a rate-limited app
Reddit -
89
Two valid tool results can describe a state that never existed
Reddit -
90
-
91
My nightly agent-learning job did nothing for a week and logged success every night
Reddit -
92
Data Table Extraction for POs and Reconciliation: How I Measure Accuracy [Workflow Included]
Reddit -
93
What was the first thing that broke when your agent went from demo to production?
Reddit -
94
-
95
-
96
What is something that you have successfully automated with an AI Agent?
Reddit -
97
-
98
-
99
What does your agent use for web data, and what have you had to hack around?
Reddit -
100
Would these automations be more easily achiveable through Loop/Planner or Lists?
Reddit