|
Enjoying this? Subscribe to get it in your inbox →
|
|
Solo Stack
|
|
Wed·Aug 12·2026
|
|
|
|
|
Copilot adds a cheaper vision coding model
A solo developer can now route lightweight coding and image-understanding tasks through MAI-Code-1.1-Flash in GitHub Copilot instead of burning premium requests by default. GitHub says Microsoft’s new small-tier coding model adds native vision support, improves coding quality and tool use, and carries a 73% lower list price than MAI-Code-1-Flash. Annual Copilot subscribers see it billed at a 0.25× premium request multiplier. The practical use case is simple: reserve expensive models for hard architecture work, and use cheaper fast models for routine edits, screenshots and small fixes.
|
|
|
JetBrains Copilot now remembers project context
A one-person technical business loses time when every coding session starts with the same project explanation. GitHub’s JetBrains Copilot update adds persistent memory across agent chat sessions, Ollama as a bring-your-own-key local model provider, and enterprise controls for plugin availability, MCP access, permission bypass behavior and OpenTelemetry. It also restores file and folder references in Copilot, Claude and Codex chat inputs. For a solo builder, the useful shift is continuity: the assistant can carry stable project facts forward while local models handle work that should stay closer to the machine.
|
|
|
OpenAI brings Daybreak cyber models to AWS
A technical founder running customer data or production software should treat security automation as part of the stack, not a later enterprise luxury. OpenAI says Daybreak Blue and Daybreak Red are now available through Amazon Bedrock for eligible customers, supporting defensive security work, vulnerability research, exploit validation and incident response inside existing AWS environments. This is not a mass-market solo tool yet, but it shows where security workflows are headed: governed model access, familiar cloud controls and faster movement from vulnerability discovery to a validated fix.
|
|
|
|
|
|
|
Audit model spend before changing tools
Before cancelling or upgrading an AI tool, check which models actually consume the budget. GitHub’s new AI usage report breaks Copilot usage down by model, including input, output, cache read and cache write tokens beside the AI credits consumed. That gives a solo operator a cleaner diagnostic than a vague monthly bill. If one workflow is expensive, the fix may be narrower prompts, better caching, a cheaper model or fewer retries. Cost control should start with the usage pattern, not with a blanket downgrade.
|
|
|
Buy AI capacity by role, not ego
If a business account needs more AI capacity, assign it where work actually bottlenecks. OpenAI’s upcoming ChatGPT Business Premium seats cost $125 per user per month, or $100 annually, versus Standard seats at $25 monthly or $20 annually. Premium gives 5× more usage and removes the five-hour usage limit. The useful lesson is allocation discipline: the founder, builder or operator doing heavy analysis may need more capacity, while occasional users stay standard. Seat mixing beats buying the top tier for everyone.
|
|
|
Make finance workflows traceable before automating them
Do not automate finance work until the decision path is visible. OpenAI’s finance-team writeup argues for redesigning workflows around the decision, then connecting source data, approvals, exceptions and ownership. For a solopreneur, that means starting with one recurring financial decision: pricing, ad spend, inventory, cash runway or tax preparation. Map the inputs and the required sign-off before adding AI summaries or dashboards. The goal is not faster spreadsheets; it is a traceable operating view where AI prepares the evidence and the owner keeps judgment.
|
|
|
|
|
|
|
ChatGPT ads expand into five more markets
OpenAI says ChatGPT Ads launched on August 11 in the United Kingdom, Mexico, Brazil, Japan and South Korea after earlier pilots in the U.S., Canada, Australia and New Zealand. Ads remain limited to logged-in adult users on Free and Go tiers, with Plus, Pro, Business, Enterprise and Education excluded. For solopreneurs, the signal is commercial, not just product-related: conversational search is becoming an ad channel. The takeaway is to watch how buyers ask for recommendations, because discovery may move from search results pages into assistant conversations.
|
|
|
Copilot web adds clearer conversation controls
GitHub made Copilot Chat on github.com easier to keep in the flow. The August 10 update adds recent-conversation access, lets users minimize the chat window while waiting for a response, and adds token-spend indicators showing per-session and per-message quota. These are small interface changes, but they matter for a solo developer using GitHub as the operating console for code, issues and reviews. The practical takeaway is that AI work should stay visible and resumable; hidden conversations and invisible spend create avoidable friction.
|
|
|
GitHub sets MAI-Code model migration deadline
GitHub will deprecate MAI-Code-1-Flash across all Copilot experiences on September 10, 2026, and recommends MAI-Code-1.1-Flash as the replacement. Copilot Enterprise administrators may need to enable the new model policy before teams can select it. For a solo business using Copilot directly, the disruption should be minor; for anyone maintaining repeatable coding workflows, model names and defaults still need housekeeping. The takeaway is to review saved prompts, scripts and team instructions before the old model disappears.
|
|
|
Solo Stack · tools · tips · news
|
|
|