New open-source tool turns any photo into a full 3D world in five minutes
image-blaster just dropped on GitHub and it does exactly what it sounds like: drop in one photo, get back a full 3D world.
Here is what comes out the other side:
- Explorable 3D environment with real depth and geometry you can navigate
- Individual object meshes with physics colliders, exported as .glb or .obj files
- Ambient sound and object-specific sound effects, auto-generated
- Works directly in Unity, Unreal, Godot, Blender, and Three.js
Under the hood, it runs as a Claude Code skill, meaning Claude acts as the agent that coordinates everything. It picks which objects to make 3D, removes them from the background, renders them with Hunyuan 3D, and adds audio via ElevenLabs. All in under 5 minutes.
To try it: git clone the repo, drop an image into the input folder, give Claude your World Labs and FAL API keys, and ask it to blast. Each provider has free trial credits to start.
Open-source tool lets Claude agents code your repos inside Minecraft
AgentCraft turns your coding workflow into a Minecraft studio you can literally walk around in. Forget staring at a terminal. Your AI teammates exist as characters in a 3D world.
Here is how it flows. You type a goal. Marlow (the lead agent) breaks it into tasks. Worker agents then code in parallel, each in their own isolated copy of your repo so they never conflict. When an agent hits a decision only you should make, it physically walks over and asks you in-game.
The safety model is solid:
- Nothing is ever pushed. Agents have zero git network access.
- Risky commands trigger an in-game permission prompt showing exactly what gets allowed.
- You review the real code diff inside the game, then hit merge.
To get started you need Minecraft Java Edition, Java 25, Node 22+, and a Claude API key. Point it at any existing repo and go. Fully open-source, MIT licensed.
Aleph Alpha open-sources 78B model with only 3.5B active params and 1M token context
Aleph Alpha just dropped Kolibri-1, a European open-weight reasoning model you can run on your own hardware. The weights are yours under Apache 2.0, so no usage restrictions.
Here is what makes it interesting architecturally: it uses a Mixture-of-Experts design, meaning the model has 78B total parameters but only activates 3.46B per token. Think of it like a team of specialists where only the relevant ones show up for each task. You get big-model quality at small-model inference cost.
What you can actually build with it:
- Process up to 1M tokens of context, great for large documents or long codebases
- Agentic tool calling with reasoning mode at low, medium, or high effort levels
- Strong German and English support, purpose-built for both languages
- RAG pipelines, coding, and multi-step reasoning tasks
To run it, install the aleph-alpha-inference package and serve via vLLM. It exposes an OpenAI-compatible API, so you can drop it into existing setups. Minimum hardware is 2x A100 80GB or 1x H200.
SpaceX Launches Google’s AI Chips; Costs Loom
Space is a harsh place to debug a server.
The Google Project Suncatcher satellite reached orbit on Oct. 1 aboard a SpaceX rocket. Google has confirmed contact with the Planet-built prototype. As we noted last week, its four AI chips match one terrestrial server’s computing power—so don’t cancel your cloud bill yet. It’s a test flight, not an open data center.
Once commissioned, Google will fire the chips in 15-minute bursts to avoid frying them while testing radiation and heat tolerance. Cooling is tricky with zero airflow. Two 2027 satellites will test laser links to sync spacecraft as a computing cluster.
Near-constant sunlight offers expanding AI capacity without taxing Earth’s grids, but getting there isn’t cheap. In Google’s research paper, one modeled path to viability requires 1,800 Starship launches carrying 200 metric tons apiece to push launch costs under $200/kilogram by 2035. At that rate, lifetime launch costs approach ground power spend. That’s a conditional projection, not a rocket order.
Google proved it can get hardware to orbit; now it must prove the hardware can do real work. Cooling, laser links, and a rocket bill that survives a spreadsheet, and managing an orbital infrastructure layer stand between this box and an orbital server farm.
Trump answered the slowdown crowd with a Super Intelligence Force
On Saturday Trump stood up a task force called the Super Intelligence Force and put Jay Clayton, the sitting director of national intelligence, in the chair. It has 120 days to deliver a report on the risks and opportunities of advanced AI. The mandate language is the tell: develop plans for responding to “SI-enabled threats to our society, while preventing overregulation and regulatory capture that would stifle innovation and competition.”
The vice chairs are FTC chair Andrew Ferguson, OPM director Scott Kupor, and Emil Michael, the Pentagon’s undersecretary for research and engineering. Condoleezza Rice and former AI czar David Sacks sit as outside advisers. The whole thing runs on the vocabulary from last month’s executive order telling agencies to write “super intelligence” instead of “artificial intelligence.”
Look again at the first vice chair. Ferguson’s FTC is reportedly drafting civil investigative demands for OpenAI, Anthropic and METR over autonomous-agent risk. He now also sits on the body charged with preventing overregulation. That is not a contradiction so much as a signal about which of the two jobs is the priority.
Clayton has been blunt about the slowdown proposals now circulating: “I don’t think any American should think that that’s a good strategy.” Four months of study, with the conclusion pre-announced.
OpenAI’s safety report writer quit and took it to The Atlantic
David Robinson, who led the writing of the safety reports that ship alongside OpenAI’s major launches and was among the company’s longest-tenured employees at three and a half years, resigned and published an essay in The Atlantic saying the “culture is broken.” His argument is structural, not personal: OpenAI “has thrived by trial and error (which it calls ‘iterative deployment’),” an approach that “guarantees periodic failures.”
His fix is that frontier labs should run “like nuclear-power plants or busy airports, with layers of redundancy and careful, time-consuming planning.” The line that should sting: in three and a half years he “never encountered a colleague who had experience making airplanes fly safely or nuclear reactors run without melting down.” OpenAI spokesperson Drew Pusateri said the company is “making sure our models don’t become more capable than we can safely manage,” pointing to security work and the paused training runs.
A day later Altman sat for Politico’s Decoded and drew the line himself. On Dario Amodei: “I think there’s a lot of daylight [between us].” On the trade: he would not take a deal where “we’ll make sure there’s no major hacks… zero… bad things.” On concentrated control by a single lab in San Francisco: a “completely unacceptable trade-off.” He wants lighter-touch regulation with targeted safeguards reserved for catastrophic risk.
Robinson’s exit follows safety head Johannes Heidecke’s departure, and Jacob Coxon, who worked at both OpenAI and Anthropic before resigning, now says the companies are “gambling with our lives.” Two frontier labs, two public philosophies, and the people who wrote the safety documents keep leaving.
Micron Sees RAM Shortages Worsening Through 2028
The shortage has excellent margins if you’re selling it.
If you were hoping for cheaper RAM anytime soon, brace yourself. Micron warns that memory and storage demand will vastly outstrip supply in 2027 and 2028, making the current squeeze look mild by comparison. The chipmaker has already locked in buyers for over three-quarters of its projected 2027 production, shoving most new contract negotiations into 2028. CEO Sanjay Mehrotra noted the company currently has no clear timeline for when the market will actually catch up to demand.
The insatiable appetite for AI infrastructure is actively cannibalizing standard component availability. Because AI is the tech world’s golden child right now, manufacturers are prioritizing highly lucrative high-bandwidth memory (HBM) for AI accelerators. This eats up factory floor space previously dedicated to conventional DRAM.
Even though overall DRAM and NAND output will technically increase, Micron anticipates the industry will continuously fall short of global demand. And while new fabrication plants are under construction, it takes years for production to meaningfully scale up once the doors finally open. Apparently, the solution to a waiting list just has its own waiting list.
While IT buyers weep, Micron is wiping away its tears with record-breaking stacks of cash. The company posted $54.23 billion in quarterly revenue alongside an 87% adjusted gross margin, nearly doubling its margin from a year earlier. It projects $61.5 billion in revenue for the coming quarter, with an adjusted gross margin hovering around 86.25%. That’s quite a view from the supplier side of the squeeze.
For IT teams, this multiyear crunch means hardware refreshes will require earlier budgeting and painful compromises over which machines truly deserve a RAM bump. Consumers eyeing new PCs could also face higher price tags or leaner out-of-the-box memory configurations. Naturally, these are just Micron’s projections; prices and availability can still shift as market demand and production capacities evolve.
Your next RAM upgrade may need a budget upgrade first.
OpenAI is putting ads inside image generation
OpenAI announced this morning a new visual ad format and said it will start testing it during image generation in ChatGPT, in the US, later this month. The ads show “product inspiration, product usage, or the experiences they make possible,” labeled and kept separate from user content. The audience OpenAI cites for the surface: 1.2 billion people each week.
The rest of the post is plumbing, which is the real news. Data-platform integrations with Hightouch, Tealium and LiveRamp. Attribution partners including AppsFlyer, Adjust, Branch, Kochava and Triple Whale. Geo-based causal experiments with Haus, Measured and WorkMagic. Brand-suitability pilots with DoubleVerify and Integral Ad Science, which OpenAI says grade the ad environment without reading private conversations. One performance number made the post: WeightWatchers’ attributed cost per acquisition on ChatGPT Ads came in 15.3% below its blended paid-search benchmark.
OpenAI’s guardrail sentence is “advertising does not influence the answers ChatGPT provides.” Image generation is a clever place to test that claim, because a product photo is native to the output rather than bolted onto an answer. No pricing, and no user opt-out, was published.
Main Street America doesn’t want to pay for AI — but that may not matter
Wall Street can’t stop talking about AI’s economic impact, but Main Street still isn’t convinced it’s worth $20 a month:
The subscription gap: Just 2.2% of US households paid for an AI service as of April 2026, according to venture capital firm a16z. While that’s roughly double 2025’s figure, it’s still a surprisingly small sliver.
This suggests a disconnect. Big Tech is citing record demand for AI and building new capacity at a rate outpacing the largest infrastructure projects in US history. But almost none of that demand is coming from households. This chart has racked up 4M views and ignited a debate over Main Street’s slow adoption:
- Microsoft’s Nicolas Bustamante suggests we’re just early in the adoption cycle and most people underestimate what today’s AI tools are capable of.
- Others have pushed back, arguing they just don’t need AI’s help to write emails, make reservations, or book flights (the use cases most commonly shown in product demos).
Either way, it’s clear where AI’s demand is coming from
Enterprises and power users who do pay for AI can’t get enough. According to a16z, the top 1% of users outspend the rest of the top 10% by roughly 8x. And when prices fall, they don’t pocket the savings — they usually buy more.
Black Forest Labs opens up image generation in FLUX 3
FLUX 3 Image uses bounding boxes to offer more precise control over your image’s layout. Just draw a box for each element, describe what should go in them, and the model renders a final image around your layout. This is a broader release of BFL’s multimodal FLUX 3 model, which originally launched in July.
Micron announces record revenue, citing continued AI demand
The chipmaker’s quarterly revenue hit $54B, up nearly 380% from $11B last year, as demand for AI continues to drive memory sales. Micron has already stopped selling RAM to consumers this year to prioritize AI customers, and CEO Sanjay Mehrotra expects memory demand to continue outpacing supply through 2028.
ChatGPT’s finance tool expands to more users
Finances in ChatGPT, first announced in May, is now rolling out to Free and Go users in the US. The feature lets you connect your bank or investment accounts so ChatGPT can answer questions about your money, track spending, and build custom financial plans. Since launching the tool, OpenAI has added credit monitoring, stock watchlists, and weekly updates.
How to measure your brand’s AI search visibility with Conductor AI
Step 1: Open Conductor’s AI Visibility Analysis tool.
Step 2: Enter your company website and work email, then click ‘Submit’.
Step 3: Open the email from Conductor in your inbox and click ‘Access Your Report’.
Step 4: Review your AI Visibility Report to see how your brand appears across AI search, including:
- AI Market Share: See your brand mentions vs. competitors.
- Funnel Visibility: Know where you appear with top prompts by search intent.
- Your Top Content: See the exact pages on your site AI is already citing.
- Your Brand Sentiment: Understand how AI perceives your brand (positive vs. negative).
Optional: Open ‘Ask Conductor’ to dig deeper into your results.
Sample Prompt: “Where does my brand currently stand in AI search, where are my biggest visibility opportunities, and which pieces of content are performing best? Identify the strongest and weakest parts of my AI visibility, and explain what may be driving those results.”
Europe’s Sovereign AI Speaks Chinese
Europe is building its sovereign AI on Chinese open-weights models. OpenEuroLLM, the EU’s sovereign-AI project spanning about 20 institutions, just fine-tuned Alibaba’s Qwen3.5-9B on Finland’s LUMI supercomputer into a model speaking 24 European languages, scores up from 59.3% to 68.1%. Its roadmap names a distillation strategy. All this while OpenAI and Anthropic keep accusing Chinese labs of stealing their outputs.
Watch the loop run its lap. American frontier gets distilled in Hangzhou, published free on Hugging Face, fine-tuned in Finland, and rebranded as European sovereignty. Even Mistral now resells GLM 5.2 from China’s Z.ai. Washington threatens bans at every summit. Nobody listens, because everyone in the loop is paid: China buys EU influence with free weights, Europe buys a sovereign stack for the cost of a fine-tune, and Nvidia sells the GPUs either way.
When interests run this deep, bans stay rhetorical on purpose. The threats were never meant to stop the loop. They are its soundtrack.
Musk Kisses the Ring
President Trump has pushed the federal government to stop saying Artificial Intelligence and start saying Super Intelligence, or SI. Elon Musk echoed it on X: “No more AI, SI, It’s Better.” On October 4, he went further and said he will rename SpaceXAI to SpaceXSI. “Yes, we will make that change,” he replied. The rename is so far only a promise. The X account still reads SpaceXAI and no timeline exists.
Musk’s empire increasingly hangs on AI infrastructure: Colossus 2, the Memphis datacenter headed from 550,000 GPUs to 1.2 million by December. The bet lives mostly on regulatory goodwill, and the government is stepping deeper into steering AI. Vocabulary is the newest lever. Nothing compelled this rename. He pre-complied. A man betting on datacenters does not risk spelling it wrong.
Musk is renaming his company to flatter Trump. The flattery is free. The datacenters are not.
Sam Altman’s Safety Was a Speech
Sam Altman has spent years telling lawmakers OpenAI puts safety first. “If this technology goes wrong, it can go quite wrong,” he warned Congress. The man who ran that safety just resigned. David Robinson lasted three and a half years, wrote the Preparedness Framework, and signed off on 12 frontier launches. His farewell essay in The Atlantic says the culture is broken: move fast, fix it later, on machines that may not offer a later.
The play behind the speech: safety reviews written, graded, and applauded by the same shippers they clear. Real safety, he says, looks like a nuclear plant. Humans are assumed to fail, and the walls catch them. OpenAI hired none of them: not one colleague had kept a reactor from melting. There was no time to fix the culture, only time to fix the model. The safety talk was for Congress. The sprint was for the product.
Altman has warned the world for years about what AI could do. His safety chief just warned it about OpenAI. Only one of them had to resign to say it.
Amazon Puts a Price on Trust
America is in the biggest data center buildout in history, and the deals ran on secrecy: tech companies had county officials sign NDAs barring them from telling residents who was building, or what the project would do to local water and power. Amazon CEO Matt Garman just announced the company no longer uses NDAs, pledged $1 billion for host communities, and blamed the backlash on foreign-seeded “misinformation and outright lies.” Experts find no evidence of that.
The move is about speed, not sunlight. Moratoriums are the one cost no capex plan survives: every county that votes no takes a campus off the table. So Amazon swapped secrecy for charm, treating goodwill as a permitting accelerant. The math is generous, too. Amazon put $13 billion into its South Bend campus alone, making five years of community sweeteners a rounding error on a single project.
Permission is the new land, and Amazon just posted the going rate.