Sign in

Bruno

@lucattelli.bsky.social
91 followers 284 following 258 posts

software engineer from brazil tilde.club/~lucattelli

PostsRepliesMedia
Bruno @lucattelli.bsky.social · 26/09/2026
+1
000
Bruno @lucattelli.bsky.social · 25/09/2026
Software fundamentals matter more as AI scales: concurrency, local state, message passing, cancellation. This paper applies them to LLM workers: a 0.6B model solves 72% of 25×25 Sudoku puzzles; a larger model roughly halves document QA latency with similar accuracy. arxiv.org/abs/2607.01077
arxiv.org
Message Passing Enables Efficient Reasoning
While inference-time scaling has improved the reasoning abilities of large language models (LLMs), the need to generate long chains-of-thought (CoTs) is a computational bottleneck. Thus, in contrast t...
121
Bruno @lucattelli.bsky.social · 25/09/2026
Andrew Ng's team went through 10k+ job posts and boiled AI engineering down to 4 skills: shipping AI apps, solid software fundamentals, knowing how to drive coding agents, and having enough product sense to decide what to build. www.deeplearning.ai/the-batch/th...
deeplearning.ai
Andrew Ng and DeepLearning.AI Present the Essential Skills Every AI Engineer Needs in 2026
I am delighted to present The AI Engineering Skills Map. AI allows us to build software very differently today than in 2022, and everyone with...
110
Bruno @lucattelli.bsky.social · 25/09/2026
Trump orders AI rebrand as "super intelligence". HAHAHAHAHAHAHAHAHA www.axios.com/2026/09/22/t...
010
Bruno @lucattelli.bsky.social · 25/09/2026
Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. github.com/Graphify-Lab...
github.com
GitHub - Graphify-Labs/graphify: Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: l...
Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local deterministic AST parsing...
000
Bruno @lucattelli.bsky.social · 25/09/2026
Why He Built Zig, Banned AI Contributions, and Moved Zig off GitHub www.infoq.com/news/2026/09...
infoq.com
Andrew Kelley Interview: Why He Built Zig, Banned AI Contributions, and Moved Zig off GitHub
In a JetBrains interview, Zig creator Andrew Kelley expanded on the project’s formal ban on AI contributions and the motivation behind its migration off GitHub to Codeberg: automated code submissions ...
000
Bruno @lucattelli.bsky.social · 25/09/2026
Agent to agent communication (A2A) is coming. benhylak.substack.com/p/agent-to-a...
benhylak.substack.com
Agent to Agent communication is coming. How will it work?
Your privacy will depend on the social skills of your agents.
000
Bruno @lucattelli.bsky.social · 25/09/2026
"Fully autonomous RSI is not happening today, and we should not pursue it unless and until it can be done safely." openai.com/index/buildi...
openai.com
Building standards for the next phase of AI
OpenAI outlines a path to shared global AI standards, calling for coordinated evaluation, reporting, and governance to improve safety.
010
Bruno @lucattelli.bsky.social · 25/09/2026
www.together.ai/blog/how-to-...
together.ai
How to train your own Jev for $17
We just launched our own Jev-like classifier, together/Tev1-4B-experimental, on top of Qwen3.5 4B on Together’s serverless platform. In this blog post we’ll show you how to fine-tune your own version!
000
Bruno @lucattelli.bsky.social · 25/09/2026
github.com/Q00/ouroboros
github.com
GitHub - Q00/ouroboros: Agent OS: the agent gets smarter on its own. We just hold the line: Interview-gated, staged evaluation, budgeted evolution loop. MCP server, 14 runtimes: Claude Code, Codex CLI...
Agent OS: the agent gets smarter on its own. We just hold the line: Interview-gated, staged evaluation, budgeted evolution loop. MCP server, 14 runtimes: Claude Code, Codex CLI, Gemini CLI, OpenCod...
000
Bruno @lucattelli.bsky.social · 25/09/2026
github.com/Human-Agent-...
github.com
GitHub - Human-Agent-Society/reef: Continual learning infra for self-improving agents
Continual learning infra for self-improving agents - Human-Agent-Society/reef
000
Bruno @lucattelli.bsky.social · 25/09/2026
Wanna play with that github.com/apowerb/apow...
github.com
GitHub - apowerb/apowerb: The open-source agentic framework to build, orchestrate, and operate production AI agents.
The open-source agentic framework to build, orchestrate, and operate production AI agents. - apowerb/apowerb
000
Bruno @lucattelli.bsky.social · 25/09/2026
github.com/confident-ai...
github.com
GitHub - confident-ai/deepteam: DeepTeam is a framework to red team LLMs and AI agents.
DeepTeam is a framework to red team LLMs and AI agents. - confident-ai/deepteam
000
Bruno @lucattelli.bsky.social · 25/09/2026
github.com/superdesignd...
github.com
GitHub - superdesigndev/treg: OpenRouter for agent tools. Join community here: https://discord.gg/6mQYYfFMAn
OpenRouter for agent tools. Join community here: https://discord.gg/6mQYYfFMAn - superdesigndev/treg
000
Bruno @lucattelli.bsky.social · 25/09/2026
github.com/jaredpalmer/...
github.com
GitHub - jaredpalmer/kev: Jev-like family of decision models built on top of Qwen3.5/3.8 you can train and run on your own
Jev-like family of decision models built on top of Qwen3.5/3.8 you can train and run on your own - jaredpalmer/kev
010
Bruno @lucattelli.bsky.social · 25/09/2026
bend-lang.com
bend-lang.com
Bend
Bend: a fast language that blocks AI mistakes via proof.
000
Bruno @lucattelli.bsky.social · 25/09/2026
www.uber.com/us/en/blog/p...
uber.com
How Uber Protects Against Retry Storms
What if retries could know when they’re helping, and when they’re making things worse? Uber Engineering built error ownership to prevent and contain retry storms.
000
Bruno @lucattelli.bsky.social · 24/09/2026
I've worked as a contractor/consultant for many years. This is SO true! Companies would rewrite/replace my software often cause they didn't understand it how I used to. www.youtube.com/watch?v=A7WF...
youtube.com
ZTA: Zero Token Architecture - Kelsey Hightower | PlatformCon 2026
YouTube video by Platform Engineering
001
Reposted by Bruno
Luiza Jarovsky, PhD @luizajarovsky.bsky.social · 24/09/2026
Meta never seems to get tired of targeting children. It announced a dedicated device to use its new AI agent, and SURPRISE, it looks exactly like a Tamagotchi, a 1990s toy. Children will start playing with it, and Meta will say that wasn't its intention. Take note.
143
Bruno @lucattelli.bsky.social · 08/09/2026
If the frontier model labs are concerned (and not fully able to contain the impact of their models), the average insurance brokerage, chemical company, or law firm needs a very strict plan to safely use AI www.bbc.com/news/article...
bbc.com
OpenAI chief scientist warns no-one is prepared for consequences of AI
The post comes as the firm releases GPT-6 Astra, which it says is its most powerful product yet.
000
Bruno @lucattelli.bsky.social · 08/09/2026
www.bbc.com/news/article...
bbc.com
Graduate job vacancies drop by almost 50% in a year, survey suggests
Job site Adzuna said the number of graduate vacancies was the lowest since 2016
000
Bruno @lucattelli.bsky.social · 07/09/2026
papers.ssrn.com/sol3/papers....
papers.ssrn.com
AI Slop
<div> The term “AI slop” has become popular to describe the output of generative AI systems seen as voluminous, low quality, or the result of little effort. Wh
000
Reposted by Bruno
OpenAlternative @openalternative.co · 05/09/2026
Setting up a workflow orchestrator usually means a database first, DAGs later. Dagu flips that: it's a single binary, DAGs are plain YAML, and it runs jobs locally or over SSH with retries and a web UI, no database required. New on OpenAlternative: openalternative.co/dagu
openalternative.co
Dagu: Open Source Alternative to Google Cloud Composer, AWS Step Functions and Astronomer
Single-binary workflow orchestrator that runs scheduled DAGs locally or over SSH, with retries, human tasks, a built-in Web UI, and file-based state storage.
072
Bruno @lucattelli.bsky.social · 04/09/2026
radio.omarchy.org
radio.omarchy.org
Omarchy Radio
Omarchy Radio
000
Bruno @lucattelli.bsky.social · 03/09/2026
Fable write me a C course that takes me from my 25yo C class to the latest and greatest o C today. 2h later I have a 37 chapter course with estimated time to completion of 2 years. Sounds like it will cover only the basics.
100
Reposted by Bruno
TechCrunch @techcrunch.com · 02/09/2026
Uber is laying off 10% of staff, or 3,300 people
techcrunch.com
Uber is laying off 10% of staff, or 3,300 people
Uber is laying off about 3,300 people, or about 10% of its global headcount, in a bid to reduce management layers and invest more in its ride-sharing, delivery, and robotaxi divisions.
02310
Reposted by Bruno
Liam @ GamingOnLinux @gamingonlinux.com · 29/08/2026
Debian Linux developers vote to allow "Responsible Use of Generative AI" #AI #Debian #OpenSource #Linux #GenAI
gamingonlinux.com
Debian Linux developers vote to allow "Responsible Use of Generative AI"
The flood gates for generative AI in Debian Linux are now seemingly open, with the results from a recent vote now available.
78911
Bruno @lucattelli.bsky.social · 29/08/2026
go.dev/blog/generic...
go.dev
Generic Methods - The Go Programming Language
Go 1.27 adds generic methods—a highly desired language feature.
000
Bruno @lucattelli.bsky.social · 29/08/2026
Yeah, right. Charging who's trying to follow the rules is a great incentive. In practice this is economically closing a door for companies trying to legally bring workers. www.insidehighered.com/news/quick-t...
insidehighered.com
Trump Administration Proposes New H-1B Visa Fee
After a federal court struck down the Trump administration’s effort to levy a $100,000 fee on new H-1B visa applications, the Department of Homeland Security is trying a different approach to impose a...
000
Bruno @lucattelli.bsky.social · 29/08/2026
uwaterloo.ca/news/media/m...
uwaterloo.ca
Major security weaknesses found in leading open AI models | Waterloo News | University of Waterloo
Story about a Waterloo-led study which found all open-weight AI models tested could be modified to bypass safety protections, raising concerns about misuse.
000
Bruno @lucattelli.bsky.social · 29/08/2026
www.theregister.com/on-prem/2026...
theregister.com
US datacenters tripled their water footprint in 10 years
... and those are figures from the start of the AI boom. It can only be worse now. Silo-ed reporting isn't helping
000
Bruno @lucattelli.bsky.social · 29/08/2026
apnews.com/article/meta...
apnews.com
Meta reaches landmark $18 billion settlement with states in trial over teen social media addiction
Meta has agreed to pay up to $18 billion and add stronger child-safety measures to its Facebook and Instagram platforms as part of a landmark legal settlement.
000
Reposted by Bruno
Kelsey Hightower @kelseyhightower.com · 27/08/2026
People don't hire AI agents, nor do people hire mouse and keyboards. AI agents are software tools; not people.
2015224
Reposted by Bruno
The Verge @theverge.com · 25/08/2026
Parents who want to keep tabs on their kids taking an Uber ride can now view a live feed of the trip through the driver’s selfie camera. www.theverge.com/transportati...
theverge.com
Uber is letting parents check in on their teens through drivers’ selfie cameras
Uber says privacy is a top priority.
1163
Reposted by Bruno
Luiza Jarovsky, PhD @luizajarovsky.bsky.social · 21/08/2026
There is no plan in AI. There is no person, company, or government that knows how to govern it in an effective, robust, and predictable way. There is currently no law, safeguard, mechanism, or magic switch that can guarantee that no AI will get out of control. Zero. Zilch.
1104
Bruno @lucattelli.bsky.social · 16/08/2026
In my experience Go gives you performance closer to system programming languages while gives you AI support closer to JS/Python. Main problem is keeping code human-maintainable as you use more. developers.googleblog.com/why-go-is-an...
developers.googleblog.com
Why Go is an Ideal Language for AI-Assisted Software Engineering- Google Developers Blog
As AI shifts software engineering from writing to reviewing, discover how Go's strict compiler and unified toolchain ensure reliable AI-generated code.
010
Bruno @lucattelli.bsky.social · 16/08/2026
I'll leave the opinion to you techcrunch.com/2026/08/13/i...
techcrunch.com
In a first, US will allow some private firms to carry out cyberattacks | TechCrunch
The new order sweeps away decades of existing U.S. cybersecurity policy prohibiting private companies from conducting 'hack back' attacks or offensive cyber operations.
000
Reposted by Bruno
Olivia Guest · Ολίβια Γκεστ @olivia.science · 14/05/2026
Yes, it's not formally finite. In other words it can contain things like an infinite loop. And as much as it is tough to wrap your head around, here's a piece of hardware that if not manually stopped would destroy itself because of such a situation. Computing is really cool, eh? youtu.be/JU9ICaPZUCg
youtu.be
What happens when you divide by zero on mechanical calculator
YouTube video by Impaler The Vlad
44512
Reposted by Bruno
Luiza Jarovsky, PhD @luizajarovsky.bsky.social · 14/08/2026
Correct, Harold
041
Reposted by Bruno
Sam Harsimony @harsimony.bsky.social · 14/08/2026
Qwen3.8-27B is out! On the benchmarks they included, is close to or exceeds Opus 4.6 huggingface.co/Qwen/Qwen3.8...
huggingface.co
Qwen/Qwen3.8-27B · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
2295
Reposted by Bruno
Tom @moll.dev · 14/08/2026
I largely agree with Bender here, we shouldn't over anthropomorphize models. But I fear that things like "prompt" and "answer" are already in the vernacular, and it's already difficult enough to vulgarize AI (disam.) terminology. buttondown.com/maiht3k/arch...
buttondown.com
How to talk about "AI" without adding to the anthropomorphization
Emily M. Bender and Nanna Inie In our op-ed for Tech Policy Press ("We Need to Talk About How We Talk About 'AI'"), we made the case against the...
232
Reposted by Bruno
Freya Holmér @freya.bsky.social · 14/08/2026
cats
318112
Reposted by Bruno
ProPublica @propublica.org · 14/08/2026
A study concluded that the uninhibited pumping of groundwater by farmers, cities and corporations around the world now accounts for 68% of the total loss of fresh water at the latitudes where most people live. (Published July 2025)
propublica.org
“Staggering” Water Loss Driven by Groundwater Mining Poses Global Threat
A new study finds that freshwater resources are rapidly disappearing, creating arid “mega” regions and causing sea levels to rise.
30752437
Reposted by Bruno
Sonja Drimmer @sonjadrimmer.bsky.social · 13/08/2026
Writing means doing a shit-ton of research, then writing, then thinking you're saying something, realizing you've approached the limit of your knowledge on a sub-topic, going away to read more, realizing you actually mean something else better or more specific, then doing it all over again.
417741
Reposted by Bruno
mr. TIM @timkellogg.me · 13/08/2026
DeepSeek Harness v0.1 the idea is everything is a plugin — models, tools, sessions, skills, sandboxes, loops, … github.com/deepseek-ai/...
github.com
GitHub - deepseek-ai/deepseek-harness: DeepSeek Harness: Everything is a Plugin.
DeepSeek Harness: Everything is a Plugin. Contribute to deepseek-ai/deepseek-harness development by creating an account on GitHub.
7669
Bruno @lucattelli.bsky.social · 13/08/2026
Pagerduty is not the best kind of alarm clock to wake up to.
100
Reposted by Bruno
jbz @jbzfn.bsky.social · 13/08/2026
🔭 The MOTHRA Telescope Captured the Faint Aftermath of a Star’s Death www.nytimes.com/2026/08/12/s... #science #astronomy #mothra
nytimes.com
https://www.nytimes.com/2026/08/12/science/mothra-telescope-helix-nebula.html
011
Reposted by Bruno
Kate Watson @loreandordure.com · 13/08/2026
Now we have the full data, the illumination curve covering the full period of the 95% eclipse here yesterday.
A light intensity curve in W/m2 showing the partial eclipse bringing the illumination down to nearly 0 before recovering while sunset continued.
1224
Reposted by Bruno
Luiza Jarovsky, PhD @luizajarovsky.bsky.social · 13/08/2026
A new type of health advice just dropped:
0105
Reposted by Bruno
Luiza Jarovsky, PhD @luizajarovsky.bsky.social · 12/08/2026
This Canadian politician forgot to delete the AI prompt from his speech:
111