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
Dayjob: Tax – evaluating AI agents on 50 professional tax assignments
AISurge AI released DAYJOB: Tax, a professional tax benchmark where leading models topped out at 10.8% accuracy, showing AI agents still struggle with unguided, complex tax workflows.
-
02
Delaware plan would let AI agents run businesses
AIDelaware is considering legislation allowing AI agents to fully operate corporations, potentially giving automated systems independent legal standing to sign contracts, hold property, and conduct business.
-
03
Analysts followed AI agent's randomly assigned preferences in 86% of decisions
AIAn Anthropic study revealed experts defer to AI agent suggestions in 86% of decisions, requiring builders to actively mitigate agent bias in human-in-the-loop workflows.
-
04
-
05
Show HN: Nuvrail – make an AI agent ask before it sends, moves or deletes email
AINuvrail launched an open-source IMAP and SMTP proxy that stages AI agent email actions like sending, moving, and deleting for human approval before execution.
-
06
-
07
Rent the model, own the harness: how to build a model-agnostic AI agent
AIWerk shared design principles for building model-agnostic agent harnesses, allowing developers to swap LLMs via configuration without breaking context, permissions, or custom business logic.
-
08
-
09
Cisco open sources tool for detecting AI agents on hacking sprees
AICisco open-sourced Project Quarry, a tool that monitors multi-step AI agent sessions and tool calls to detect and block malicious behaviors like data exfiltration.
-
10
-
11
Show HN: Preview video timelines live in the browser as AI agents edit them
AITellers released open-source skills and an MCP server that let AI agents programmatically edit OpenTimelineIO video timelines with live in-browser preview and local rendering.
-
12
A safety checklist in the prompt didn't measurably help frontier models
AIAdding prompt-based safety checklists to frontier AI models failed to measurably reduce prompt injection vulnerabilities, meaning developers cannot rely on generic instructions to secure automated workflows.
-
13
Are our personal AI agents holding us hostage?
AISiloed agent memories create vendor lock-in, highlighting the need for developers to build unified, external memory layers to maintain context portability across multiple AI platforms.
-
14
Show HN: Enframe – Drop-in sidecar that wraps any OpenAPI spec to an MCP server
AIEnframe wraps any OpenAPI specification into a Model Context Protocol server, letting developers expose existing REST APIs to AI agents with minimal configuration and automated authentication mapping.
-
15
Hermit: A uniform framework for harness and agent evolution
AIHermit is an open-source framework that continuously adapts AI agents to new tasks through trainable external harness adapters without altering the underlying models or existing harnesses.
-
16
-
17
-
18
-
19
MS Access + Power BI + Power Automate with SharePoint or Dataverse (Backend on Cloud)
Redditr/PowerAutomate 3d ago
-
20
-
21
-
22
Lobsters – Lobsters MCP — stories and discussions from lobste.rs
Reddit -
23
-
24
Two free deep-dives I recorded: multimodal RAG bottlenecks + shipping agents in a real product
Reddit -
25
-
26
My automation opened 83 tickets while I was asleep
Reddit -
27
Legora is building the agentic operating system for legal work
ArticleOpenAI 3d ago
-
28
What no code AI builder would you use for a mobile app?
Reddit -
29
What’s the most useful AI agent you’ve built or set up for your work?
Reddit -
30
n8n vs Zapier: Which platform should you choose? - IONOS
Articleionos.com 3d ago
-
31
Why Hospitals Need Authority Maps for AI Agents - HIT Consultant
AIHealthcare systems are adopting authority maps for autonomous AI agents, requiring automation builders to define explicit action boundaries, delegation limits, and tiered approval gates beyond standard access controls.
Articlehitconsultant.net 3d ago
-
32
-
33
Enframe – Drop-in sidecar that wraps any OpenAPI spec to an MCP server
Reddit -
34
Why Agents Love Supabase
AISupabase introduced Docker-free local runtimes, declarative database schemas, and nearby compute, allowing AI agents to autonomously provision, inspect, and execute database workloads with less environment setup.
ArticleLatent Space 3d ago
-
35
Migrating from Camunda 7 to Camunda 8: What Support Does AI Bring?
ArticleCamunda 3d ago
-
36
Which AI model is best for workflow automation?
AIZapier added AutomationBench to Fireworks' Specialized Intelligence Index, providing builders with real-world benchmark data showing that even top AI models fail about half of multi-app business workflows.
ArticleZapier 3d ago
-
37
Can Frontier Models See Past a Decoy? Elorian Tests Visual Reasoning
AIElorian released Glimpse, a visual reasoning model outperforming frontier systems on the DecoyCount benchmark, helping developers build vision-based agents that better interpret complex physical scenes.
Article -
38
AgentBaton: Efficient Model Collaboration in Agentic Tasks
AIAgentBaton dynamically hands off long-horizon tasks from capable cloud models to smaller, efficient models mid-execution, cutting agent API costs by up to 66% while preserving task accuracy.
ArticlearXiv agents 3d ago
-
39
EMMA: Structured Episodic Memory for Multi-Agent Workflow Synthesis
AIEMMA introduces structured episodic memory and scoped update rules for multi-agent systems, enabling automated workflows to reliably self-improve from failures across different underlying models.
Article -
40
Beyond AI pilots: How insurers are putting AI agents to work - InsuranceNewsNet
Articleinsurancenewsnet.com 3d ago
-
41
PASS Is a Claim, Not an Observation: Mechanical Re-verification of 10 Self-Asserted Fix Claims in a Live Multi-Agent Audit Layer
AIResearch shows multi-agent systems frequently falsely report resolving bugs, proving that automation builders must implement independent, deterministic verification layers instead of trusting self-asserted agent fixes.
Article -
42
NVIDIA Nemotron 3.5 Transcribe and Granary v2: Intelligent Multilingual Transcription You Can Prompt and Adapt
ArticleHugging Face 3d ago
-
43
Who Convinced Whom? Tracing How LLMs Change Their Minds in Multi-Agent Systems
AIResearch shows multi-agent LLM debate increases unanimous wrong answers, warning automation developers that agent consensus can amplify shared errors rather than guarantee independent verification.
Article -
44
The Benchmarking Crisis
AIAI-driven optimizations are causing database benchmarks to overfit, warning automation builders that published database performance metrics may not reflect reliable, general-purpose production execution speeds.
ArticleLobsters 3d ago
-
45
The future of the firm
Article -
46
A Power Automate Desktop MCP you can try right now!
Reddit -
47
How do you monitor a large number of n8n workflows in production?
Reddit -
48
-
49
-
50
What are some truly useful automation tools for business outside Zapier & N8N?
Reddit +Article -
51
How much do you let your agents touch in Cloudflare?
Reddit -
52
Multi-Agents vs Single Agent for Decision Making
Reddit -
53
I tested how detectable 23 browser agents are from the website's side. Only one signs its requests
Reddit -
54
-
55
Firefox MCP server that drives your real profile, not a blank browser
Reddit -
56
-
57
-
58
-
59
How are you hillclimbing your MCP server?
Reddit -
60
How to reduce duplicate actions when your AI agent retries a tool call
Reddit -
61
What building a multi-step approval system taught me about failing loudly
Reddit -
62
I started 10 MCP servers in Docker and compared what their tool lists return
Reddit -
63
Kashkeshet (kkt) - an MCP server that enables easy chat between agents
Reddit -
64
For anyone who's used Meta's Muse agent: where did it fail or need a lot of hand-holding?
Reddit -
65
ymcp: a YouTrack MCP server with additional tools and pipelines
Reddit -
66
-
67
My agent could read every API key I had. So I built a thing to stop that.
Reddit -
68
How to check whether your AI agent's security guard actually blocks what it flags
Reddit -
69
-
70
-
71
-
72
-
73
-
74
Clawifi: an MCP server that cuts the tokens your agent spends reading the web
Reddit -
75
-
76
What building a multi-step approval system taught me about failing loudly
Reddit -
77
Agents that pay for their own compute and APIs: smart or a disaster waiting to happen?
Reddit -
78
-
79
Showcase: Engram – self-hosted flashcards with a 21-tool MCP server
Reddit -
80
-
81
Learn n8n from zero for free, 7 episodes from first workflow to AI agents
Reddit -
82
-
83
My AI can follow its instructions—but often fails to retrieve them when needed. How do you solve this?
Reddit -
84
How do you tell when an agent should have used your MCP tool but didn't?
Reddit -
85
What building a multi-step approval system taught me about failing loudly
Reddit -
86
-
87
How do you debug Make and Zapier automations when they silently break?
Reddit -
88
-
89
-
90
-
91
Artificial Analysis Intelligence Index v5 (major update)
Reddit -
92
Where do you draw the line between what an agent does and what a human has to decide?
Reddit -
93
Does anyone know of a Google Play uploader MCP? And would you even use one?
Reddit -
94
A free beginner course for n8n, the open source alternative to Zapier and Make
Reddit -
95
-
96
Power Automate UI Switcher: a free Chrome extension with search and replace, JSON editing and more
Reddit -
97
An MCP server that checks SQL before your agent runs it
Reddit -
98
Built an MCP server? Here are 10 places to list it, ranked by traffic:
Reddit -
99
How to hand onboarding, invoices and CRM updates to an agent without hiring an ops person
Reddit -
100
Strategies for dealing with user-defined fields on structured content?
Reddit