frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Ask HN: What are the biggest limitations of agentic AI in real-world workflows?

2•aadarshkumaredu•1h ago
There’s a lot of momentum around agentic AI systems that can plan and execute multi-step workflows autonomously.

For teams that have tried deploying these in production environments, where do they actually break down?

Is it reliability over long action chains, tool integration issues, cost unpredictability, state management, latency, observability, or something else entirely?

I’m especially interested in failure modes that only became obvious after moving beyond controlled demos into real usage.

Comments

buschleague•1h ago
State management. The agents lose track of what they already did, re-implement things, or contradict decisions from 20 minutes ago. You need external state that survives compaction because the agent can't be trusted to maintain its own.

Constraint adherence degrades over long chains. You can put rules in system prompts, but agents follow them for the first few steps, then gradually drift. Instructions are suggestions. The longer the chain, the more they're ignored.

Cost unpredictability is real but solvable.

Ultimately, the systems need external enforcement rather than internal instruction. Markdown rules, or jinja templates etc., that the agent can read (and ignore) don't work at production scale. We ended up solving this by building Python enforcement gates that block task completion until acceptance criteria are verified, tests pass, and architecture limits are met. The core learning being that agents can't bypass what they don't control.

Study: Self-generated Agent Skills are useless

https://arxiv.org/abs/2602.12670
1•mustaphah•38s ago•0 comments

ID Token Nicer

https://anddata.substack.com/p/id-token-nicer
1•allusernamesare•1m ago•0 comments

Show HN: Wildex – we built Pokémon Go for real wildlife

https://apps.apple.com/us/app/wildex-identify-plants-animals/id6748092158
3•AnujNayyar•4m ago•0 comments

Show HN: Free Alternative to Wispr Flow, Superwhisper, and Monologue

https://github.com/zachlatta/freeflow
1•zachlatta•5m ago•0 comments

Data 101 for Business Leaders (Operator-Focused, No Mumba-Jumba Jargon)

https://www.datadrip.com/blog/why-were-launching-data-101-and-why-most-companies-misunderstand-data
1•datadripsol•6m ago•1 comments

Nothing bad is going to happen – junkie [video]

https://www.youtube.com/watch?v=_N4z1H6csSs&list=RD_N4z1H6csSs
1•marysminefnuf•6m ago•0 comments

Npx Build-Skill

https://www.npmjs.com/package/build-skill
1•flashbrew•7m ago•1 comments

Show HN: Twsla – A tiny, high-speed log analyzer written in Go

https://github.com/twsnmp/twsla
1•twsnmp•11m ago•2 comments

Obituary for Robert Duvall

https://www.theguardian.com/film/2026/feb/16/robert-duvall-actor-passion-conviction
3•Archelaos•15m ago•0 comments

Has NASA set another prelaunch test after delay? Latest on Artemis

https://www.usatoday.com/story/news/nation/2026/02/16/nasa-artemis-2-rocket-launch/88646242007/
2•bookmtn•16m ago•0 comments

Baby #2, no job, + tradeoffs and prioritizing

https://www.sonyasupposedly.com/baby-2-no-job-tradeoffs-prioritizing/
3•exolymph•18m ago•0 comments

Claude Cowork

https://pvieito.com/2026/01/inside-claude-cowork
3•tin7in•19m ago•0 comments

The One Woman Anthropic Trusts to Teach AI Morals

https://www.wsj.com/tech/ai/anthropic-amanda-askell-philosopher-ai-3c031883
2•jdkee•21m ago•1 comments

Visualize the entropy of a code base with a 3D force-directed graph

3•Piprim•22m ago•0 comments

Show HN: Your binary is no longer safe

https://reorchestrate.com/posts/your-binary-is-no-longer-safe/
2•seddonm1•23m ago•0 comments

Claude 4 Sonnet: Conversation with Kai

https://docs.google.com/document/d/12woq_BpFbzLkH4zHvVRJLPyiZGoDVDxA/edit
2•doener•24m ago•0 comments

Show HN: Provisioner per-board sidecar for serial access, flashing, and bring-up

1•acarminati•24m ago•0 comments

A Pocket Guide to Surviving the Robot Apocalypse

https://corbt.com/posts/a-pocket-guide-to-surviving-the-robot-apocalypse/
1•kcorbitt•26m ago•0 comments

Virtual Width Networks (VWN)

https://arxiv.org/abs/2511.11238
5•tesserato•26m ago•1 comments

Kung Fu Robots Deliver 'Knockout' Performance at Spring Festival Gala [video]

https://www.youtube.com/watch?v=NVX6vq0RSnY
2•zaikunzhang•26m ago•0 comments

Show HN: Scrappy – Open-source browser scraper written in Go

https://github.com/sunergos-ro/scrappy
4•RaulOnRails•26m ago•1 comments

User "Claude" committing vulnerabilities at a rapid rate

https://cyberplace.social/@GossiTheDog/116080909947754833
3•not2b•26m ago•0 comments

Java has evolved. Your code can too

https://javaevolved.github.io
1•pjmlp•27m ago•0 comments

The left is missing out on AI

https://www.transformernews.ai/p/the-left-is-missing-out-on-ai-sanders-doctorow-bender-bores
2•themgt•27m ago•0 comments

UK government may 'age restrict or limit children's VPN use'

https://www.techradar.com/vpn/vpn-privacy-security/uk-government-says-it-may-age-restrict-or-limi...
1•iamnothere•30m ago•0 comments

AI Rot Management

https://jw.hn/ai-code-hygiene
1•jwpapi•34m ago•0 comments

Big tech stocks lose billions as AI spending fears hit valuations

https://www.reuters.com/business/retail-consumer/global-markets-marketcap-2026-02-16/
5•billybuckwheat•35m ago•0 comments

Robert Duvall, a Chameleon of an Actor Onscreen and Onstage, Dies at 95

https://www.nytimes.com/2026/02/16/movies/robert-duvall-dead.html
3•bookofjoe•36m ago•1 comments

Vulkan Easy => Reactor (Like React-App)

https://github.com/AndreeSalazar/REACTOR-Framework-for-Vulkan-
2•QDanteX•36m ago•0 comments

Anthropic got an 11% user boost from its OpenAI-bashing Super Bowl ad

https://www.cnbc.com/2026/02/13/anthropic-open-ai-super-bowl-ads.html
2•general_reveal•36m ago•0 comments