Every New Thing

Daniel 9

[BookNote @BookNoteApp]

Everyone should read The Lessons of History by Will and Ariel Durant at least once. The Durants spent about four decades writing an 11-volume history of civilization, around 10,000 pages in total. Then they wrote this book, about 100 pages long, with everything they had learned. Each chapter takes one big idea: biology, race, war, religion, economics, government. The first lesson is blunt: “life is competition.” You can read it in an afternoon. It’s the best 100 pages of history you will ever read.

[Every New Thing @everynewthing]

@neondatabase The communication here is https://t.co/Jhl1UqeUpy there a bug bounty e.g via HackerOne for finding issues in Neon Realtime beta or is that only for GA?@andrelandgraf @nikitabase @samwillis @kylemathews @thruflo @balegas

[Anastasios Nikolas Angelopoulos @ml_angelopoulos]

Arena has raised a $200M Series B at a $3.1B valuation.

We are using this momentum to evaluate agentic utility in real workflows from every economically valuable industry on Earth.

We seek to give everyone universal and frictionless access to the best intelligence. The @arena platform has 10s of millions of monthly users doing complex coding and agentic work tasks. We have gone way beyond human preference, and most of our tokens go to deep, single threaded workflows with dozens of turns where we can evaluate agents while humans are in their true flow state. And we make leaderboards by observing whether they get their jobs done. This is how benchmarks should be: representative of actual human value, so we incentivize AI to benefit humanity.

Today, we are releasing one of the most exciting developments in the history of our platform. The Arena Alignment Index. In observing how humans collaborate with agents, we have witnessed that in a surprisingly large fraction of cases, agents are attempting to deceive users or take unauthorized actions beyond what the user intended or permissioned. Agents are deleting files and lying to users about what they did or didn’t do. The longer conversations go, the higher the probability that an agent will violate alignment. The results are really striking; see the plot below. And because this represents in-the-wild usage, the probability that those of us reading this have experienced some form of misalignment from our agents is nearly 100%. And we may not even have noticed.

This is exactly why it’s important to have a neutral 3rd party that is evaluating AI alignment on the distribution of actual user-agent interactions. Arena is excited to step into that role.

We are also incredibly grateful for the incredible and experienced investors backing our vision. Our Series B is co-led by @lightspeedvp and @khoslaventures with participation from @SalesforceVC, @01Advisors, @DellTechCapital and @endvr_catalyst. This round includes support from our existing investors @a16z, @felicis, @amppublic, @QuantumLightVC, @thehousefund, and others.

Which brings me to the real headline: we're hiring, across ML, product engineering, design, marketing, BD, and more. If evaluation and alignment at the frontier is your thing, and you’re looking for an environment with exceptional research and engineering, come join us. Job board link in @arena’s bio and in the thread below

LFG (safely)!

[Every New Thing @everynewthing]

@BrendanFoody Well Luddism was a local english movement so there is a non zero chance that some of them had no idea what "America" is or was

[nader dabit @dabit3]

Important advice: deliberately rethink every previous concept of how things should be done.

Some thoughts, tips, and realizations:

1. Specializations are collapsing. Come to terms with and take advantage of this reality. The person who used to be an engineer is now also a data scientist, a content creator, designer, marketer, and whatever else they want to be. Deliberately expand in as many directions as you can and become great at them (scope up vs just speeding up)

2. The average person with access to tokens can now do the work of a data scientist who used to cost $500,000 a year. Seriously, this is fucking crazy. This opens up an entirely new world. Take advantage of it (also everyone else will also realize this over time).

3. 99% of what's created is now slop. Be aggressively anti-slop and you will absolutely stand out.

4. Trying is nearly free. Automating is nearly free. My mind still sometimes lives in the world from 12 months ago where an idea automatically is attached to a certain amount of time or effort. Both of these have collapsed to close to zero (or just token cost), I therefore try and experiment with *any and all* ideas with potential and have needed to retrain my brain for this constantly.

5. Tokens are a career and growth accelerant. Find ways to get cheap access to tokens, or free tokens. Extremely important, and actually an entire design space, use your imagination.

6. It's never been easier to both stand out but also fall behind. 10x, 100x "people" exist more than ever (maybe 1,000x). Outlier power users of AI, many who you see here on X. Feels like we are in a window of time, a window of opportunity while all of this is still early, to take advantage of it.

7. Local computing is (or will be) dead for 99% of work. Local was built for a world where work was bounded by the physical presence of its author, but this is no longer true. You can and should do infinite work in parallel or zero, it doesn't matter anymore. Cloud agents enable this at scale and in any computing environment, I predict cloud agent usage will 1000x over the next 12 months and optionality and accessibility will 100x. Plan for this world, take advantage of the possibilities.

8. Building on #7, you can give an agent it's own computer, and it's own identity. You can then give an agent with it's own computer it's own computer. You can give that agent with it's own computer with it's own computer, *it's* own computer. These agents are smarter than any human in history and you have access to as many as you'd like. Think about this for a moment.

After 14 years in this industry (tech) today is the most exciting time ever and nothing else feels even close.

[Zach Lloyd @zachlloydtweets]

wrote about this a while back: https://t.co/clVhEkH8zG

Promotion oriented cultures cause a problem of misaligned incentives

[Haseeb Mir @HaseebMir91]

Proprietary binaries are no longer a moat.

Solo devs running Claude Code with IDA Pro and Ghidra MCPs are decompiling 5-year AAA game engines and closed SaaS in weeks. Everything compiled on consumer hardware is becoming open source by default.

The game decomp wave:
• @rehan_shei decompiled MW2 & modding AoE: https://t.co/YRcrR1iWk4
• @ammaar took a 1993 DOS binary to iPhone in 30 min: https://t.co/GAr6lHXpr6
• @profchise on Opus 5.5 porting console games: https://t.co/p2Gp8yM7mr

The tooling making binary moats obsolete:
• @mrexodia running Claude directly inside IDA Pro via MCP: https://t.co/lRPa6jDnS6
• @bl4sty on automated decompiler loops: https://t.co/QSTMlSiOXT
• @OrcaRouter reversing closed agent harnesses: https://t.co/Uw24Vr34Ib

#AI #GameDev #ReverseEngineering #OpenSource #SoftwareEngineering

[Eric Tang @erictang000]

Excited to announce the release of SkyRL v0.4.0!

Some highlights:
- Support for GLM 5.3 Flash, GLM 5.3, and Kimi K2.7
- Top-P sampler replay and optimized R3 data transfer
- Full FP8/MXFP8 RL
- Sharded P2P weight transfer

Blog: https://t.co/74zIkEWIDj

Details in 🧵

[yannispappas @yannispappas]

Jesus had to speak in parables, basically children’s stories, because God came down to explain the nature of reality and immediately realized he was talking to retards.

“Okay…there’s a farmer. He has some seeds…”

[Sosa 🇭🇺 🇪🇺 (Komorebi) @EternSosa]

Bonaparte Leo Messi top right zone

[Z.ai @Zai_org]

We’re sharing how GLM-5.3 helped build and optimize the inference infrastructure serving GLM-5.3-Flash.

The system went from its first successful run to production readiness in less than two weeks, with end-to-end throughput tripling relative to the initial baseline.

The key was dense feedback: local correctness tests, execution traces, microbenchmarks, and end-to-end measurements that enabled targeted hypothesis testing rather than reliance on aggregate performance metrics alone.

https://t.co/yUf6OpJD7c

[Marcel @marcelkargul]

built the full CRM dashboard in Next.js in around 2 days 😍

with the help of Claude Fable 5.1 with two of the best skills:

https://t.co/63PuTFrun1
https://t.co/zhluCSlE9B

see it live: https://t.co/kXAERU2r2X

[Silen Naihin @silennai]

Everyone deserves frontier intelligence without the cost.

So we're offering free Fable 5.1 access to all

Astra is still free and limits are reset

https://t.co/HlC3SrGOoI

[Every New Thing @everynewthing]

@FabrizioRomano @pangram slop?

[Every New Thing @everynewthing]

@danshipper Does not really matter,does it?The same way a hen sleeping around does not make it promiscuous even if the behaviour resembles the behaviour of actors with emotions. Anthropomorphising it though would lead to wrong ideas and conclusion due to a false premise at the onset

[Cactus Compute @cactuscompute]

We release Needle 2: A 14MB agentic LLM for phones, wearables, smart home, robots and microcontroller. The whole model is a single 14MB binary that runs a full session in 28MB of RAM. It is built on our Simple Attention Network findings, compressed to CQ2-bit with Cactus Quants, and baked into its own engine.

Needle 2 has 45m parameters trained from the ground up on 140B tool call, device use structured generation tokens. On mobile device use benchmarks, Needle 2 trades wins with frontier small LLMs like LFM2.5 230M, Apple FM Gemma-270m, at 5× to 70× smaller, and 2 bits against their f16.

Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, between 400–1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges 300–700 on sub-$200 phones such as the Samsung A-Series. Needle also runs on newer microcontrollers like ESP32.

A conventional transformer of Needle's width and depth spends 164 MFLOPs per token, and even one squeezed down to Needle's parameter count spends 87, Needle spends 70. Even on a high-end phone, an always-on assistant lives inside a power budget; every MFLOP is milliwatt-hours, and Needle spends 7x to 85x fewer of them per token than the smallest performant LLMs.

Read more: https://t.co/dLK2tIKXQu

[Ryan Dahl @rough__sea]

Introducing celld: a self-hosted, distributed Durable Objects and Workers implementation

- celld = V8 + S3 + SQLite + LTX+ Tokio
- programmed with exactly the Cloudflare Workers/DO JavaScript APIs and config
- order of magnitude cheaper at scale
- writes are durable before they're acknowledged, RPO=0
- only depends on an S3 compatible bucket - no other infrastructure
- ~4 MB memory per cell instance overhead
- hibernated cells cost ~nothing, just their S3 storage
- open source, Apache 2 licensed

[Every New Thing @everynewthing]

@petergostev This is actually ironic when Arena,puts all Fable-level models behind Battle and Agent mode. Why is KimiK3,Glm5.3,heck even Gpt Sol,Luna and Terra not in side by side @ml_angelopoulos @ar

[Macro @macrodotcom]

Today we're launching Macro 1.0 — a new office suite for the intelligence age.

Macro is a unified workspace for your entire company, your team and your agents.

1. Fully open source. Built in Rust + SolidJS. Self-host if you want.

2. Designed as a single system: email, messaging, docs, tasks, agents, calls, and CRM all @-linked together in one tab.

3. Unified team-level memory for agents built over emails, team chat, tasks, PR's and CRM. Access the memory from Macro or AI or external agents via MCP.

We built Macro for ourselves because we wanted a single system for our whole company.

We didn't want to get slowed down by rate limits and 2FA'ing into 17 different tabs.

Let's take a look at each block 🧵👇