Open Source Factory

POLICY & RESEARCH Gemini 4 Argon: A Long Output Needs a Budget Open Source Factory · Sep 30, 2026 · 5 min read A million-token output ceiling lets a generation continue for a long time. It does not tell you how much of the result will be usable. Argon's launch deserves a budget for accepted work, alongside the separate questions of API rates and initial access. In this article Output length and the context budget The coding rows Argon does not lead The initial access program Pricing a long run Availability at launch Image: Google. Official launch artwork; the following charts are our own redraws. Output length and the context budget Google raises its advertised output ceiling from 64K to 1M tokens. That describes how far a generation can run; it is separate from the input context window. Official output-limit announcement . I find the distinction more interesting than another giant-context headline. Reading a large repository and sustaining a long attempt to modify it sol...

HARDWARE & DESIGN Ray-Ban Meta Audio or Gen 3: Pay for a Camera You'll Use Open Source Factory · Sep 30, 2026 · 6 min read The starting-price gap is $100. For that premium, Gen 3 needs a capture job that Audio cannot do for you. Listening and calls make a different case, especially for someone who prefers glasses without a camera. Start with the use, then compare the lens quote and frame fit. In this article The camera premium Battery claims for Audio Privacy beyond the camera Putting a frequency on capture Frame, lenses and delivery checks My starting recommendation is Audio for someone who mainly wants listening and calls. Gen 3 makes more sense when hands-free capture is already part of the plan. This is a buying analysis based on the announcements, not a hands-on review; I have not tested either pair. Audio in the Clubmaster style at Meta Connect. Photo: Nicholas Sutrich / Android Central. The pictured lens configuration is not a $349 price quote. The camera premium Me...

AGENTS & WORKFLOWS OpenAI DevDay 2026: What Could You Use at the Event? Open Source Factory · Sep 29, 2026 · 4 min read DevDay's agent story spans models, tools and shared work. For a developer, the useful distinction is which parts were available at the event and which remained previews or promises. The September 29 roundup below keeps those rollout states beside the features. In this article Persistent work between prompts The developer-tool changes Shared work inside ChatGPT Live features, previews and promises Some features are live, some depend on the plan or market, and others remain previews or announcements. The sections below separate those states so a developer can tell what can be tested at the date of the event. A self-rendered concept image for the shared-workspace theme; not an OpenAI product screenshot. A grouping of the announcements by function, based on OpenAI’s official recap . Persistent work between prompts Dots is the clearest sign of that direction...

AGENTS & WORKFLOWS Two $10 Coding Plans: What One Month's Usage Shows Open Source Factory · Sep 29, 2026 · 7 min read Neither plan interrupted the month of work recorded here. That is useful evidence about this workload, not a promise of unlimited use. The meter readings, models and length of the agent runs explain why a higher tier was difficult to justify for this particular month. In this article The workload behind the meter readings Plan limits side by side Differences that remain below the cap Models outside the $10 plans The $40 tier and the community’s objections Linked source documents Not once did I hit a limit on either plan. A typical 30-minute agent run — DeepSeek V4.1 Flash chewing through a real program while I did something else — burned about 2% of GOAT's 5-hour window and about 3% of OpenCode Go's. Total spend across both plans: roughly $21 including tax, with zero top-ups. So when OpenCode launched the $40 Go Plus tier on September 28, my first hon...

MODELS & COSTS Claude Sonnet 5.5: Choose the Effort Level With the Model Open Source Factory · Sep 28, 2026 · 12 min read Anthropic's launch card for Sonnet 5.5, shipped Monday, September 28. (image: Anthropic) Sonnet's lower token rate does not make every setting cheaper than Opus. The independent effort ladder changes that comparison at the expensive end. For repeated work, choose a model and reasoning setting together, then compare the cost of accepted results. In this article Release and migration details Anthropic’s scorecard and its settings The independent index Cost at different effort levels Cyber safeguards and fallback frequency Responses from developers and users Choosing a setting for the workload This release follows the Opus 5.5 comparison, but the useful question is specific to Sonnet: how does the effort setting change the price of a completed task? The launch documents and independent index give different views of that cost. The cyber-fallback footn...