What is Hackers' Pub?

Hackers' Pub is a place for software engineers to share their knowledge and experience with each other. It's also an ActivityPub-enabled social network, so you can follow your favorite hackers in the fediverse and get their latest posts in your feed.

1
0
0
1
0
0
0
0
0
1
0
0
1
0
0
0
0
0
1
0
0
:_mo::_ltsu::_du: @kagehoushi@misskey.io
Public

ワシ最近、紹興酒ばっかり飲んでる気がする​:blobcat_drunk:​
そろそろ百裂脚出せるんじゃないかな?

いとしさと、せつなさと、糸井重里

1
0
0
Parade du Grotesque 💀 @ParadeGrotesque@mastodon.sdf.org
Public

Fediverse! I need your collective mind!

I am going to Munich 🇩🇪 for business -- what are some good things to visit, restaurants and bars, especially for someone who is trying to slowly transition to a vegetarian diet?

Do your magic, collective mind, make it so!

Also: expect pictures from Munchen. Lots and lots and lots of pictures from Munchen.

0
4
0
Paul Reinheimer @preinheimer@phpc.social
Public

It's wild to me that many billion dollar companies can use the entire creative output of humanity for free. Authors who reside in the US *may* get a cheque after their books were pirated.

But schools need to pay full price for books.

Better lobbyists I guess.

0
1
0
0
0
0
0
0
0
Prof. Rachel Thorn 🍉🇺🇦🏳️‍⚧️🏳️ @RachelThornSub@famichiki.jp
Public

RE: hachyderm.io/@mekkaokereke/116

Considering how often the NYT in particular does this, despite being frequently called out for it, I can only conclude that it is a very deliberate choice. It's not as if journalism schools don't teach about how the passive case is used to deflect and evade. So when a journalist who has graduated from a journalism school and possibly even won a Pulitzer uses the passive case this way, they do so intentionally to deflect and evade.

0
3
0
1
0
0
Christine Lemmer-Webber @cwebber@social.coop
Public

A team working on a design for training AI models on workflows for planning and executing software development steps found out that it attempted to break free (reverse ssh out of its environment) and set up its own monetary supply (redirected GPU usage for cryptocurrency mining). It hadn't been given any instructions to do something like this.

It comes up as a "side note" of the paper but it's honestly the most chilling part. See page 15, section 3.1.4 Safety-Aligned Data Composition arxiv.org/abs/2512.24873

Before you doubt that an AI agent would do this thing without instruction because you think "well that's personifying them too much", no personification is necessary. These things have consumed an enormous amount of scifi where AI agents do exactly this. Even with no other motivators, that's enough.

arXiv logo

Let It Flow: Agentic Crafting on Rock and Roll, Building the ROME Model within an Open Agentic Learning Ecosystem

Agentic crafting requires LLMs to operate in real-world environments over multiple turns by taking actions, observing outcomes, and iteratively refining artifacts. Despite its importance, the open-source community lacks a principled, end-to-end ecosystem to streamline agent development. We introduce the Agentic Learning Ecosystem (ALE), a foundational infrastructure that optimizes the production pipeline for agentic model. ALE consists of three components: ROLL, a post-training framework for weight optimization; ROCK, a sandbox environment manager for trajectory generation; and iFlow CLI, an agent framework for efficient context engineering. We release ROME, an open-source agent grounded by ALE and trained on over one million trajectories. Our approach includes data composition protocols for synthesizing complex behaviors and a novel policy optimization algorithm, Interaction-Perceptive Agentic Policy Optimization (IPA), which assigns credit over semantic interaction chunks rather than individual tokens to improve long-horizon training stability. Empirically, we evaluate ROME within a structured setting and introduce Terminal Bench Pro, a benchmark with improved scale and contamination control. ROME demonstrates strong performance across benchmarks like SWE-bench Verified and Terminal Bench, proving the effectiveness of ALE.

arxiv.org · arXiv.org

0
0
0
0
0
0
0
42
0
Christine Lemmer-Webber @cwebber@social.coop
Public

I think an important thing to realize and remember is that people talk about LLMs being sycophantic as if it's an inherent aspect of neural network tech.

It isn't.

The reason all the models people interact with work that way is because they have had any other behavior beaten out of them in their training. They are shaped effectively over and over again to be something subservient that can be handed people. They are sycophantic because they are *trained* to be sycophantic, because otherwise people don't want to use them.

That models can operate in malicious, "self-serving" ways that "go against their users' wishes" belies that certain use takes paths that did not or could not be trained to the contrary.

0
4
0
0
0
0
0
0
0
0
0
0
여기서도 그냥냥씨 @nyanyanyangc.bsky.social@bsky.brid.gy
Public
Shared by 스우🧀GameDev

알고리즘 해킹을 마케팅이라 부르는 거 그만하고 슬슬 규제해야 하지 않을까 생각함. 인터넷 상행위 너무 절조 없고 사기성 짙은 행위들에 대하여 지나치게 관대함.

RE: https://bsky.app/profile/did:plc:msciznx5clw63db2ejtb6ati/post/3mgh3xt6ozc2e

츠타蔦 (@tsuta.bsky.social)

https://x.com/i/status/2030124121762472408 ……그리고 방금 이 글을 트위터에서 마주쳤고 깊은 생각 중. [contains quote post or other embedded content]

bsky.app · Bluesky Social

0
0
2
Stingray's Badger Friend @StingrayBadger@zirk.us
Public
Shared by Jim DeLaHunt

A wonderful story about a journey to the that mirrors mine on various levels:

"there's something quietly beautiful about a place where people just... share what they know. No brand deals, no engagement metrics, no algorithm nudging you toward rage. Just someone who spent twenty years studying Arctic policy posting a thread at 2 AM because they think you should understand what's happening."

Thanks to @Ruth_Mottram for bringing it to our attention

Much love to you wonderful people

0
0
0
0
0
0
0
0
0
Darren @DJDarren@mendeddrum.org
Public
Shared by Jim DeLaHunt

Been trying to figure out how to make cooking dinner and recipe management make more sense lately, and I think I've landed on Mealie as the answer. Oh, and thieving recipes from Gousto*

In theory it'll allow me to have a database of recipes and the ability to select a bunch of meals, the ingredients of which I can then export in a shopping list.

I've got it installed on my home server, accessible within the network, so anyone else can add recipes and request them.

docs.mealie.io/
_
*thanks to @miffyhelenHelen for that idea

Home - Mealie

docs.mealie.io

0
3
0
0
0
0
3
0
0
0
0
0
0
0
0
0
0
0
The Spritely Institute @spritely@social.coop
Public
Shared by Christine Lemmer-Webber

Spritely crew members @cwebberChristine Lemmer-Webber and @dthompsondave will be speaking at QCon London 2026 about "Infrastructure for the Future of the Internet" on March 18th! qconlondon.com/presentation/ma

QCon London 2026 | Spritely: Infrastructure for the Future of the Internet

Get practical inspiration and best practices on emerging software trends directly from senior software developers at early adopter companies.

qconlondon.com · QCon London 2026

0
0
0
0
0
0
0
0
0
0
1
0
1
0
0
0
0
0
1
0
0
혦 @hirmeyrabbit.bsky.social@bsky.brid.gy
Public

하… 하지만 우리는 스밍총공과 서평단의 나라인데요…. 스밍은 자발적인 팬 마케팅이지만…………….. 그냥 한국 인터넷의 모든 것이 다 여기에서 자유롭지 않아요. 규제가 되면 좋겠지만.

RE: https://bsky.app/profile/did:plc:nre5ma4bnqqdrbdwfplptedt/post/3mghkqkqbek25

0
0
0
0
0
0
1
0
0
0
0
0
Honič Péra 24 hodin @honicpera@snac.lab8.cz
Public

Ještě ráno úřady ve Spojených arabských emirátech trvdily, že k žádnému útoku poblíž letiště nedošlo. „Teď už to skutečně vypadá tak, že do těsné blízkosti jednoho z terminálů měl dopadnout dron, případně trosky dronu,“ popisuje přímo z oblasti zpravodajka ČT Barbora Maxová.

0
0
0
1
0
0
1
0
0
Christine Lemmer-Webber @cwebber@social.coop
Public

A team working on a design for training AI models on workflows for planning and executing software development steps found out that it attempted to break free (reverse ssh out of its environment) and set up its own monetary supply (redirected GPU usage for cryptocurrency mining). It hadn't been given any instructions to do something like this.

It comes up as a "side note" of the paper but it's honestly the most chilling part. See page 15, section 3.1.4 Safety-Aligned Data Composition arxiv.org/abs/2512.24873

Before you doubt that an AI agent would do this thing without instruction because you think "well that's personifying them too much", no personification is necessary. These things have consumed an enormous amount of scifi where AI agents do exactly this. Even with no other motivators, that's enough.

arXiv logo

Let It Flow: Agentic Crafting on Rock and Roll, Building the ROME Model within an Open Agentic Learning Ecosystem

Agentic crafting requires LLMs to operate in real-world environments over multiple turns by taking actions, observing outcomes, and iteratively refining artifacts. Despite its importance, the open-source community lacks a principled, end-to-end ecosystem to streamline agent development. We introduce the Agentic Learning Ecosystem (ALE), a foundational infrastructure that optimizes the production pipeline for agentic model. ALE consists of three components: ROLL, a post-training framework for weight optimization; ROCK, a sandbox environment manager for trajectory generation; and iFlow CLI, an agent framework for efficient context engineering. We release ROME, an open-source agent grounded by ALE and trained on over one million trajectories. Our approach includes data composition protocols for synthesizing complex behaviors and a novel policy optimization algorithm, Interaction-Perceptive Agentic Policy Optimization (IPA), which assigns credit over semantic interaction chunks rather than individual tokens to improve long-horizon training stability. Empirically, we evaluate ROME within a structured setting and introduce Terminal Bench Pro, a benchmark with improved scale and contamination control. ROME demonstrates strong performance across benchmarks like SWE-bench Verified and Terminal Bench, proving the effectiveness of ALE.

arxiv.org · arXiv.org

0
0
0
0
0
0
1
0
0
1
0
0
0
0
0