fp.
newest
Open in hackernews
Intel Accelerates Open Infrastructure for the Agentic AI Era
https://www.intel.com/content/www/us/en/newsroom/news/artificial-intelligence/intel-accelerates-open-infrastructure-agentic-ai-era.html
1
•
boyter
•
43m ago
Chip stocks slide in wake of 'casino' comment by Broadcom boss
https://www.ft.com/content/e16f0422-cf75-4cc9-962d-4605e421dc63
1
•
thm
•
2m ago
•
0 comments
CVE-2026-82456: Argo CD MCP Server Kubernetes Takeover
https://apiiro.com/blog/cve-2026-82456-argo-cd-mcp-server-kubernetes-takeover/
1
•
wondersky
•
5m ago
•
0 comments
Styx Consulting – Cybersecurity and IT Infrastructure for SMEs
https://styxconsulting.be/
1
•
Sharanxxxx
•
6m ago
•
0 comments
High School Freshman Looking to Get Better at Design
https://studio-grade.vercel.app/
1
•
arjunkshah21
•
8m ago
•
1 comments
Show HN: Scentrail, a data explorer that keeps the reasoning behind your charts
https://scentrail.io/
1
•
hyojeonglee
•
11m ago
•
0 comments
Nixpkgs for the Front End Impaired
https://fzakaria.com/2026/10/12/nixpkgs-for-the-frontend-impaired
2
•
ghuntley
•
12m ago
•
0 comments
Same Query, Different Answer: Run-to-Run Variance in Semantic Query Processing
https://arxiv.org/abs/2610.13823
1
•
sbulaev
•
13m ago
•
0 comments
How to stop your brain from rotting
https://scotto.me/blog/2026-10-12-how-to-stop-your-brain-from-rotting/
1
•
silcoon
•
18m ago
•
0 comments
Portman – Your friendly port manager on homebrew
https://github.com/arps18/portman
1
•
arps18
•
18m ago
•
0 comments
LLM Agents Can Easily Tamper with Their Own Traces
https://arxiv.org/abs/2609.30266
1
•
Kaedon
•
20m ago
•
0 comments
Things we've learned about building AI products
https://newsletter.posthog.com/p/38-things-weve-learned-about-building
1
•
swolpers
•
21m ago
•
0 comments
Grace Hopper Celebration Cancelled
https://ghc.anitab.org
2
•
uma_ch
•
22m ago
•
1 comments
They War-Gamed Dimming the Sun. It Went Worse Than You Think
https://www.wired.com/story/they-war-gamed-dimming-the-sun-it-went-even-worse-than-you-think/
1
•
Teever
•
22m ago
•
0 comments
Yaks: A tiny, fast discovery tree for software teams
https://github.com/mattwynne/yaks
1
•
mattwynne
•
26m ago
•
1 comments
To Classify, Try a Trie Tree
https://fsdatalab.github.io/blog/ai-classification/
2
•
matt_d
•
29m ago
•
0 comments
Hydropneumatic Suspension
https://en.wikipedia.org/wiki/Hydropneumatic_suspension
1
•
Svip
•
30m ago
•
0 comments
Software engineers don't get to be Luddites about AI
https://www.seangoedecke.com/software-engineers-dont-get-to-be-luddites-about-ai/
3
•
vismit2000
•
32m ago
•
0 comments
Show HN: A 3D sky city where my Claude Code agents work and take coffee breaks
https://github.com/ishraq21/skyborne
1
•
Ishraq21
•
33m ago
•
0 comments
Don't Feed the AI
https://github.com/zeroc00I/DontFeedTheAI
1
•
car
•
35m ago
•
1 comments
Show HN: GPT-3.5 answered, GPT-4 voided; I built an EOS gate for AGI
https://getswiftapi.com
1
•
rayanpal_
•
35m ago
•
0 comments
Intel Accelerates Open Infrastructure for the Agentic AI Era
https://www.intel.com/content/www/us/en/newsroom/news/artificial-intelligence/intel-accelerates-o...
1
•
boyter
•
43m ago
•
0 comments
Measure a Workflow, Not a System
https://rlafuente.com/posts/2026-10-11-agent-evals-2-measure-a-workflow.html
2
•
andes314
•
43m ago
•
0 comments
Show HN: Open Health Sync – Self-hosted API for wearable and health data
https://github.com/soma-intelligence/open-health-sync
1
•
quentincaron0
•
46m ago
•
0 comments
Why agents lose the plot on large or legacy codebases
https://loopsfinity.com/blog-context-rot-large-codebases
2
•
shymaple
•
47m ago
•
0 comments
FBI arrests Canadian cybersecurity expert in ShinyHunters hacking case
https://www.cbc.ca/news/canada/canadian-cybersecurity-expert-arrested-fbi-shinyhunters-9.7378256
2
•
empressplay
•
47m ago
•
0 comments
Costa Vayenas, Insights from "The Swiss Franc 1798 – 2055" [video]
https://www.youtube.com/watch?v=5-KdMBQkUic
2
•
johntfella
•
48m ago
•
0 comments
Fork-and-Flush: Escaping Idea Basins in Autoresearch Agents
https://arxiv.org/abs/2610.07447
2
•
matt_d
•
51m ago
•
0 comments
Stateless Language Agents: Scaling Long-Horizon Automated Research
https://arxiv.org/abs/2610.07625
2
•
matt_d
•
57m ago
•
0 comments
When should an agent own the control flow?
https://www.calm.rocks/blog/pipeline-vs-agent/
1
•
calmrocks
•
58m ago
•
1 comments
Trader Nets $1.28M on Papertrade as 1,000x Leverage Raises Risk Questions
https://coinmarketcap.com/community/post/380070089/
1
•
joeymabia1
•
59m ago
•
1 comments
Load More