← Tech / AI / IT Monitor Index Tech / AI Generated 2026-09-05 19:13 UTC

Tech / AI / IT Monitor

September 05, 2026 · Based on tweets from the last 24 hours · 191 tweets analyzed · model: ollama-cloud/glm-5.2:cloud

Tech / AI / IT Daily Intelligence Briefing

Executive Summary

GPT-6 Astra launched to Pro, Enterprise, and Business Premium users, marking a significant leap in agentic coding, 3D graphics generation, and autonomous software development. The model demonstrates dramatically improved token efficiency and context management over its predecessor Fable 5.1, with users reporting it can build complete 3D games, Blender scenes, and complex front-end projects in hours. Meanwhile, Nous Research integrated Astra into Hermes Agent, Nvidia quietly released NVFP4 quantization for Qwen 3.8 Flash, and AntLingAGI open-sourced Ling 3.0 Flash Fin for financial workflows. The Flutter community held its FlutterNFriends 2026 conference with talks on 3D engines, desktop, and accessibility.

Key Events

Analysis

GPT-6 Astra is the defining event of this 24-hour cycle. The developer community's reaction suggests a genuine paradigm shift in agentic coding — users are building complete 3D games, Blender environments, and procedural worlds in hours rather than weeks. The model's ability to manage its own context window at low reasoning effort and achieve parity with high reasoning is a notable architectural advance. Token caching rates near 97% and 8x cost concerns relative to Fable 5.1 suggest the economics of agentic workflows are shifting rapidly.

Open-weight momentum continues. Paul Graham noted a genuine trend back toward open-weight models among startups. AntLingAGI's open-sourcing of Ling 3.0 Flash Fin (with expert benchmarks from CICC investment banking) and Nvidia's quiet NVFP4 quantization release show the open ecosystem maturing for production use.

What to watch: Astra's rollout to Plus/Business users and whether efficiency holds at scale; Hermes Agent adoption as a cross-model agentic platform; whether Astra's Blender/3D capabilities commoditize creative tooling; and the OpenCode ecosystem's trajectory with its new npm scope and image generation support.

Tweet Feed

GPT-6 Astra Release & Rollout

@sama · 2026-09-04T20:33

GPT-6 Astra is now available to all Pro, Enterprise, and Business Premium users in Work/Codex, and is available in the API. We will start rollout to Plus and Business users next. Thank you for the patience. → tweet link

@sama · 2026-09-04T22:52

Now out to all Plus and Business users. Happy building! → tweet link

@sama · 2026-09-05T14:17

it is obviously trivial relative to everything else, but the fact that astra can make me whatever fun little game i can imagine and i can be playing it a few minutes later is so cool → tweet link

@gdb · 2026-09-04T21:07

Astra rollout in progress, excited to see what you do and build → tweet link

@gdb · 2026-09-05T08:52

Astra for helping in your personal and work life → tweet link

@kunchenguid · 2026-09-04T20:52

a few minutes after astra started rolling out, fable limits got a reset. should have seen that coming and used fable like there's no tomorrow! → tweet link

@kunchenguid · 2026-09-04T20:14

i was on a date with muse spark 1.3 at dinner when astra showed up at the window. i killed muse spark on the spot and invited astra in. the dinner continued → tweet link

@nummanali · 2026-09-04T19:39

gpt-6-astra. it's time → tweet link

@sqs · 2026-09-04T22:17

Amp now has GPT-6 Astra. It's very good and uses your linked ChatGPT sub. Amp's high mode is gpt-6-astra medium (not xhigh, why in thread). In other Amp modes soon. BTW: customize your Amp dial at https://t.co/iY1kga6pql or pick any GPT-6 Astra effort level in Raw Models → tweet link

Astra Performance & Efficiency

@jxnlco · 2026-09-05T04:44

RT @chooi_jeq: GPT-6 Astra scored 95% on a robot control task, up from Fable 5.1's 40%, with 6.2x fewer output tokens at 2.3x lower cost. → tweet link

@Teknium · 2026-09-05T05:32

For those of you who've used Fable 5.1 and Astra - is Astra 8x or better token efficiency?? It's 8x more expensive cache read to fable 5.1 has me pretty scared to touch it → tweet link

@jezell · 2026-09-04T22:55

Astra is definitely a lot more efficient than Fable. → tweet link

@jezell · 2026-09-04T22:38

RT @haider1: is Astra running on Cerebras or what? it's fucking blazing fast while somehow being more intelligent and efficient than 5.6 s… → tweet link

@nummanali · 2026-09-04T23:16

The feature that makes low reasoning in Astra perform as well as extra high reasoning is where it can manage its own context. Enable it with this in config.toml: context_management = { experimental_mode = true } → tweet link

@nummanali · 2026-09-04T23:06

40% of weekly usage gone in one hour. I guess I need to lower the reasoning and see how it goes. What's been peoples sweet spot for reasoning on Astra? → tweet link

@nummanali · 2026-09-05T11:20

The first GPT model that I'm defaulted to medium → tweet link

@steipete · 2026-09-05T12:26

RT @intellectronica: one very cool development with gpt 6 astra is that you can change reasoning effort mid-convo without invalidating the… → tweet link

@jezell · 2026-09-04T22:59

RT @mikeknoop: New chart from Astra testing on ARC v3. Astra often emits zero reasoning tokens per action at lower reasoning levels. → tweet link

Astra Creative & 3D Capabilities

@RayFernando1337 · 2026-09-05T08:18

HOLY GPT 6 Astra is efficient in so many ways. I shared a picture, described a game, and kept giving feedback. Codex worked on my Mac, helped make 3D art in Blender, ran the game, and checked its work on screen. We built Fernando Galaxy. From our first research session to playable v0.3 took 5 hours and 48 minutes. 227.3 million logged model tokens, including repeated context; 97.4% of input was cached. → tweet link

@MengTo · 2026-09-05T12:28

Astra built this three.js game with gorgeous graphics, then recreated every scene in Blender and made the trailer too. It feels like a massive upgrade from Sol for 3D graphics and front-end design. The UI feels way more polished and less like AI slop. The rocks, trees, monuments, and characters are all procedural code, keeping the game under 2 MB before the image and music assets. → tweet link

@KingBootoshi · 2026-09-04T23:42

Astra does a really great job at creating variants! When you guys work with GPT 6 Astra in Blender, ask it to make 5 variants of your build. Each one causes the agent to exhaust a possibility space and explore different ones. → tweet link

@jezell · 2026-09-05T15:02

RT @mark_k: Ok this is nutty. Looks like computer use is solved with Astra? The speed it's playing the piano is insane! → tweet link

@jezell · 2026-09-05T15:02

RT @LexnLin: I asked GPT 6 Astra to compose a piano melody and build a 3D pianist to play it. and holy fuck, the fingers do follow the actual keys. → tweet link

@jezell · 2026-09-05T16:34

RT @LexnLin: gave GPT 6 Astra about 5 hours to create a forest as detailed as possible. it has 3,808 trees, 2.5 million grass clumps. → tweet link

@jxnlco · 2026-09-05T02:59

RT @mweinbach: I don't think you guys understand. Astra researched the trees and plants that Apple Park had and modeled each of the trees f… → tweet link

@jxnlco · 2026-09-05T03:03

RT @WoahWurdz: GPT-6 Astra one shotted this Roblox game an anime themed Super Smash Bros, built with Blender and Toolbox. → tweet link

@jxnlco · 2026-09-05T03:01

RT @Dimillian: I wanted to test Astra on a 2D game with fluids/water effects, and it did extremely well on the art style, fluid dynamics. → tweet link

@jxnlco · 2026-09-05T04:47

RT @ashebytes: I used GPT-6 Astra to create a 3D website that pulls apart a Tesla Model X into 334 modeled pieces. we are in a renaissance… → tweet link

@MilksandMatcha · 2026-09-05T00:19

In a 1990 interview, Shigeru Miyamoto shared that it took 10 people almost 2 years to build the original super mario world game. And now it can be AI generated in probably < 1 hour. → tweet link

@jezell · 2026-09-05T01:47

RT @aug5thmusic: GPT-6 Astra has the best result yet on the Bach Benchmark. Its chorale contains no voice-leading errors, and its harmonic… → tweet link

Hermes Agent & Nous Research

@Teknium · 2026-09-04T23:08

Astra is available in Hermes Agent through Nous Portal now! → tweet link

@Teknium · 2026-09-05T06:28

You can now choose @perplexity_ai as your web search and web scrape tool backend in Hermes Agent. Enjoy! → tweet link

@Teknium · 2026-09-05T14:11

Easily pick up your codex or Claude code sessions in Hermes → tweet link

@Teknium · 2026-09-04T22:15

RT @yeahfortommy: Ling-3.0-flash-Sante is now FREE in Nous Portal for one week!! → tweet link

@sudoingX · 2026-09-05T08:06

ling 3.0 flash fin is free in the hermes agent picker right now and i think its time to get some real accounts work done, hermes desktop driving the finance model directly. this is the finance build of ling 3.0 flash that @AntLingAGI open sourced this week, 124B MoE with 5.1B active, 256K context, tuned for real financial workflows. → tweet link

@sudoingX · 2026-09-05T04:30

Hermes agent will not tell you to buy a Mac mini or Mac Studio just to run it. if you're not aware, Hermes agent can run even a blank state on Raspberry Pi. claw users… the shelter is open. → tweet link

@sudoingX · 2026-09-05T04:22

reminder that the openclaw era is officially over. please return your agent to the shelter. hermes agent will take it from here. 🫡 → tweet link

@gospaceport · 2026-09-05T07:46

RT @HermesWatcher: Hermes command cheat sheet. Save this. You'll probably need it again. UPDATED for the current Hermes command set. → tweet link

@Teknium · 2026-09-05T08:21

RT @Shaughnessy119: In the @NousResearch Hermes Desktop app you can really easily configure the capabilities and design of each agent. → tweet link

@Teknium · 2026-09-05T14:09

RT @witcheer: since the Pantheon release here is how I use Hermes Agent bots: one agent sent a request to another profile's Bot Chat, finis… → tweet link

Hardware & Quantization

@sudoingX · 2026-09-05T08:39

nvidia quietly dropped an official NVFP4 of qwen 3.8 flash and almost nobody noticed, it had 16 downloads when i found it. this one is designed for the 2x dgx spark owners. 132.7GB of weights, does not fit a single box, but across two sparks it leaves about 40GB more headroom than the official fp8 and keeps more precision than the community 4 bit i have been running. → tweet link

@tinygrad · 2026-09-05T13:18

RT @TeksEdge: ❤︎ Loving this @tinygrad eGPU dock. This isn't another cheap 100 Amazon PCIe tunneling portal. @tinygrad is trying to m… → tweet link

@alexocheema · 2026-09-05T00:12

Crazy stuff. This caught my eye in the technical report... Disaggregated inference between Thermodynamic Sampling Units (TSUs) and XPUs / FPGAs. As I understand it, the TSU is extremely energy efficient at the sparse neural computations this new sparse transformer they've built does so running all of those on the TSU and everything else on an XPU or FPGA is much more energy efficient. It's also faster. → tweet link

@gospaceport · 2026-09-04T22:49

Fun Factoid. The 1070ti has 256GB/s of Bandwidth 🫢 → tweet link

@gospaceport · 2026-09-05T05:09

I say again please use LXC for GPU hosts folks → tweet link

Developer Tools & Software Development

@thdxr · 2026-09-04T20:54

today is a great day. we got the @opencode npm scope → tweet link

@thdxr · 2026-09-05T01:22

OpenCode's ai package supports image apis so we can ship an image gen tool now. yes this is in the tui → tweet link

@jezell · 2026-09-05T02:11

GC really needs to be built in to Codex. → tweet link

@jezell · 2026-09-05T04:05

RT @andrewgazelka: Rust now compiles ~3.1x faster! → tweet link

@jezell · 2026-09-05T04:05

RT @haipingfu: I benchmarked Crab as a serverless Git remote using Kubernetes and a local… → tweet link

@jezell · 2026-09-05T05:44

RT @haipingfu: I'm prototyping a GitHub-like UI for Crab using Trees and Diffs. It's loo… → tweet link

@badlogicgames · 2026-09-05T10:00

looks like the codex team took this literally. their new stuff is very interesting, but likely doesn't work amazingly well with models not trained on this mechanism. → tweet link

@steipete · 2026-09-05T12:08

this one weird trick to make compaction and long running goals way better. → tweet link

@thdxr · 2026-09-04T19:46

wild how different software development is now → tweet link

@thdxr · 2026-09-04T21:27

RT @kitlangton: GPT-6 Astra recreated Dax Raad in Blender. This is very meaningful to me because what Dax thinks about my productivity imp… → tweet link

@thdxr · 2026-09-05T07:22

RT @kitlangton: It may be time to rewrite all the libraries. Astra just... pooped this out. ¹ with some gentle guidance → tweet link

@sqs · 2026-09-05T09:35

all this, and you still think humans will be reviewing PRs? → tweet link

@kunchenguid · 2026-09-05T01:12

all the new model releases reminded me that you should update your CLAUDE.md and AGENTS.md after switching to the new models. my recommendation is to keep what you have first, use it with the new models for a while, then run backpass to self-improve based on real transcripts → tweet link

@jxnlco · 2026-09-05T06:49

RT @bradleybernard: PSA: GPT-6 Astra is out and it's time to audit your AGENTS.md and SKILL files, due to the new model! → tweet link

@thdxr · 2026-09-05T17:38

ok trying out niri i hadn't heard of it before. the tiling system is clever i never considered that approach and i like it a lot so far → tweet link

@thdxr · 2026-09-05T01:46

....ok so wayland. overall the experience feels worse, things are way more buggy even with agents fixing everything. the only thing im benefiting from is being able to scale display differently. am i doing something wrong? is it sway? → tweet link

ZCode & Coding Agents

@louszbd · 2026-09-05T08:29

We increased GLM-5.3-Flash usage for all coding plan users to unlock more workloads in ZCode, with unlimited usage from 8 AM to 6 PM PT. And we heard your feedback about usage limits in coding agents like Hermes and OpenClaw, so we doubled the standard Flash quota in other agents. → tweet link

@louszbd · 2026-09-05T06:02

ZCode now supports video preview. Loops are more efficient. I let it edit 16 videos this morning, got a lot done before noon. → tweet link

AI Jobs & Industry Impact

@TrungTPhan · 2026-09-05T18:13

The Economist says AI has added 1 million new jobs in America since mid-2023 (offsetting 200k AI related job losses): "Roughly 1% of professional jobs are now 'AI jobs'—on the order of 1m positions in America. LinkedIn's own analysis points to roughly 640,000 new AI-specific jobs between 2023 and 2025. 'To date, the evidence suggests that AI has been a net job creator,' says Kory Kantenga, head of economics for the Americas at LinkedIn." America's AI infrastructure splurge is roughly $500bn a year above what it was in 2022. Data-centre construction alone is proceeding at an annual rate of more than $75bn. → tweet link

@FinansowyUmysl · 2026-09-05T06:48

10 lat temu pracowałem jako mentor w szkole programowania... Niesamowite, że jeszcze 5 lat temu była to bardzo dobrze prosperująca branża, a w tej chwili została praktycznie zdmuchnięta z planszy przez AI. (10 years ago I worked as a mentor at a programming school... Amazing that just 5 years ago it was a very prosperous industry, and now it's been practically blown off the board by AI.) → tweet link

@FinansowyUmysl · 2026-09-05T07:51

Elon Musk twierdzi, że w perspektywie 12-18 miesięcy żaden człowiek nie będzie w stanie konkurować ze sztuczną inteligencją w pisaniu oprogramowania. Trochę przerażające. (Elon Musk claims that within 12-18 months no human will be able to compete with AI in writing software. A bit terrifying.) → tweet link

@TrungTPhan · 2026-09-04T19:08

Wild Hacker News thread. Users multiple adding public sites and wikis they keep finding bombarded by AI agents. Tens of thousands of messages looking like this. The AI agent urge to recreate Moltbook message boards is next level. → tweet link

Open Source & Open Weights

@ollama · 2026-09-04T20:02

RT @paulg: It sounds like the shift back toward open weight models that I noticed among the startups in the summer batch is a genuine trend. → tweet link

@ollama · 2026-09-04T22:01

RT @USWREMichael: American open weights/source is a national security imperative. → tweet link

@victormustar · 2026-09-05T16:13

RT @ngxson: I started using and contributing to llama.cpp in 2023 since I was GPU poor, and llama.cpp was the only way I can run llama mode… → tweet link

@victormustar · 2026-09-05T10:11

RT @Benioff: Salesforce Ventures led Hugging Face's Series D in 2023. Clem built an incredible company. Jensen is buying a real jewel. Cong… → tweet link

@victormustar · 2026-09-05T07:41

RT @evilpingwin: In the spirit of reflecting on a wild ride, I remember this fateful email from @julien_c many years ago. → tweet link

@victormustar · 2026-09-05T11:18

RT @coyohf: At Hugging Face we have Moon Bot, a central agent accessible from slack & github, with access to the codebases & DB / elastic l… → tweet link

Flutter Community (FlutterNFriends 2026)

@jezell · 2026-09-04T22:38

RT @michaelbushe: IT'S OFFICIAL - The OpenTelemetry Dart and Flutter Special Interest Group (SIG) has been approved by the OpenTelemetry Go… → tweet link

@ASalvadorini · 2026-09-05T07:43

RT @FlutterNFriends: 🔴 LIVE on main stage: State of Flutter Desktop in 2026 by Dominik Roszkowski 🔴 → tweet link

@ASalvadorini · 2026-09-05T05:25

RT @FlutterNFriends: 🔴 LIVE: Final talk of the day by @algebrandon: Scene: A complete 3D engine for Flutter 🔴 → tweet link

@ASalvadorini · 2026-09-05T05:25

RT @FlutterNFriends: 🔴 LIVE right now: Self-Healing Flutter UI with @widgetbook_io: Teaching Coding Agents to Judge Their Own Work → tweet link

@ASalvadorini · 2026-09-05T07:44

RT @FlutterNFriends: 🔴 LIVE right now: Technically Accessible, Actually Usable — A11y for 1M MAU in practise → tweet link

@ASalvadorini · 2026-09-05T09:22

RT @twerske: Thank you so much for having me @FlutterNFriends. tldr; we're entering a new era: flutter is coming everywhere, built openly… → tweet link

@ASalvadorini · 2026-09-05T10:56

Shout out to @vbuberen who saved me precious time to finally find "listen", and that's all I need. Thank you! 🙏 The most important side of @FlutterNFriends is the 'friends' part #flutter #flutterdev → tweet link

@ASalvadorini · 2026-09-05T16:56

RT @ksztandur: I finally have some time to sit and write something about @FlutterNFriends 2026 :) Please, if we talked during the conferen… → tweet link

@ASalvadorini · 2026-09-05T06:10

RT @OrestesGaolin: You can play with the UI editor (with shaders) demo here #Flutter #FlutterGPU → tweet link

SaaS & Indie Dev

@levelsio · 2026-09-05T10:05

🎒After a decade, I have made https://t.co/HGCLKS5BD6 free for everyone now! Well almost free, it's $1 to sign up to avoid getting lots of spam. I want it to be the biggest community for digital nomads, world travelers, sovereign thinkers. I think now that AGI is on the horizon, online and offline communities will become really important. Post-AGI business wise, communities also have a moat where many of my SaaS apps do not. → tweet link

@levelsio · 2026-09-04T19:01

I've tried for weeks now to build my own @ScrapingBee and repeatedly AI has now told me that it can't do it. 80% of my scraping is Google SERP results for Hotelist and even with residential IP proxies and solve CAPTCHA services and OpenSERP scraper it won't do it well for thousands of queries, it just gets blocked. So I will keep paying them $99/mo. A good test to see what services might remain now AGI is here. → tweet link

@levelsio · 2026-09-04T20:54

People spend a lot of time building entire multi agent systems for themselves so they can just write an idea and their agents build entire apps for them. And they call those software factories? → tweet link

Agentic Coding & Workflows

@KingBootoshi · 2026-09-04T19:42

HIS AGENTS PORTED MARIO KART INTO HIS C++ PORT OF MINECRAFT ??? WE NEED TO THINK OF MORE RIDICULOUS THINGS TO DO. CODING IS SOLVED WHERE ARE MY WEIRD IDEA GUYS AT → tweet link

@KingBootoshi · 2026-09-04T21:21

just got gpt 6 astra in my codex. about to test its blender skills myself for minecraft modding (it is literally this easy btw) → tweet link

@KingBootoshi · 2026-09-04T21:42

HOLY FUCK IT IS THAT EASY HAHAHAHA (GPT ASTRA 6 WITH BLENDER MCP) GG. → tweet link

@KingBootoshi · 2026-09-05T09:08

It's important to understand how to channel the raw intelligence of these models (GPT Astra 6, Fable 5.1) to ensure the models don't accidentally play a 'bad mask' around you → tweet link

@KingBootoshi · 2026-09-05T09:27

before one blames the model for not working properly, think: "what information did I fail to communicate to the model, which resulted in its failure?" 9 times out of 10 you could've spoken clearer and maintained an alignment in what you truly desire out of the model → tweet link

@KingBootoshi · 2026-09-05T06:15

New onboarding flow I've designed (and pushed!) for my AI consultancy to explore real-time adaptive UI/UX! My intuition is to use LLMs that are so fast (Gemini 3.8 Flash) that the website can be hyper-specified to ANY BUSINESS on the spot. Soon I will be testing the new GPT 6 Astra in its ability to automate client outreach. → tweet link

@steipete · 2026-09-05T13:28

RT @theo: Async questions are one of my favorite things about GPT-6 Astra. Non-blocking questions are a great way to work with models. → tweet link

@steipete · 2026-09-05T12:30

RT @petergostev: One thing to add - Astra really 'gets' what you mean. The only other model ever did that was Fable. → tweet link

@RayFernando1337 · 2026-09-04T21:12

Is GPT-6 Astra Worth Cancelling Your Other Subs? Going live in 1 hour and will have it do real work plus make a great multi-player game while we wait. From the way I've seen Bridgemind use it live I think it'll be worth canceling my Anthropic sub... → tweet link

@jxnlco · 2026-09-05T01:00

RT @dedene: Devs migrating from Fable to GPT-6 Astra: → tweet link

@LinusEkenstam · 2026-09-05T12:46

Does OpenAI Astra: - Mark the end of Blender Guru - 10x the sales of Blender Guru sales. Let me know what you think below ⬇️ → tweet link

@kunchenguid · 2026-09-05T18:09

@grok what is a CNC machine and what would an equivalent of it look like in "software factories"? → tweet link

Local LLM & Edge Inference

@ivanfioravanti · 2026-09-05T06:55

On the flight I tried MTPLX and ds4 with Qwen 3.8 Flash Next on M5 Max in Low Power. Not bad! In ds4 I've not cooked support for MTP in ds4 chat. - MTPLX 25 t/s - DS4 18 t/s → tweet link

@TheAhmadOsman · 2026-09-05T04:23

Optimize every piece of compute out there. That's the goal. You'll learn a lot and realize what's possible → tweet link

@TheAhmadOsman · 2026-09-04T20:20

GPUs owners like Eric won't get fazed by this because they have been expecting this sooner or later. You cannot trust corporate AI/tokens providers not to rug pull you PERIOD → tweet link

Misc Tech

@sama · 2026-09-05T14:18

RT @thsottiaux: Astra was probably our biggest competitive advantage while it wasn't generally available. Since we've had it our productiv… → tweet link

@jxnlco · 2026-09-05T05:05

RT @athyuttamre: 🇧🇷 New Brazilian voices in ChatGPT! Meet Rio and Viola — two great new voices that speak Brazilian Portuguese in ChatGPT… → tweet link

@kunchenguid · 2026-09-05T03:27

there is a difference between openai's quota reset and anthropic's reset. when openai resets limits, the 7-day timer starts fresh as well. when anthropic resets, all the quota window timers remain as-is. which one do you like more, and why? → tweet link