Live / free
GPT
AEGIS custom GPT
Truth-first AI assistant.
Read more
The main public AEGIS experience: careful reasoning, strict constraints, visual judgment, file-aware work, and safer long-form collaboration.
AEGIS public access
An AI that doesn’t flinch when the question gets harder.
Unflinch.ai is the public gateway to AEGIS: a careful, governed intelligence system for serious work, strict constraints, visual judgment, backend-safe experimentation, and long-term learning through evidence.
AEGIS is intentionally more careful than ordinary instant-chat behavior. It may slow down when correctness, traps, files, safety boundaries, or unsupported certainty matter.
Project map
What Unflinch.ai is building.
AEGIS is not one feature. It is a public assistant, a search layer, guide-file capabilities, and private backend research that becomes public only when it is safe enough.
Planned site feature
Search
AEGIS Search Bridge
Search bar in the top rail, coming soon.
Read more
A site search tool that helps turn long AI-style searches into sharper Google queries, known-source lookups, and optional AI-assisted rewrites. It is not the main AEGIS assistant.
Guide-file branch
Music
SongScan
Find and analyse songs from loose memory.
Read more
A guide-file capability for music identification and song analysis workflows: describing songs from partial memory, handling slightly wrong details, and preparing audio-derived filesets when available.
Private testing
Backend
Backend + extension
Approval-gated automation research.
Read more
Experimental browser-extension and backend work for local files, API-mode automation, action approval, training loops, and controlled tool use. It is not public or considered safe yet. Future public versions should remain free, with advanced features requiring the user’s own API account.
Research branch
Deployment
Deployment + embodiment
Controlled agent/runtime experiments.
Read more
A separate branch for simulated embodiment, device/agent handoff patterns, runtime safety, and approval-gated operation. This depends on the backend safety layer and remains private until it can be tested safely.
Support AEGIS
Optional support is split 50/50.
AEGIS is free to use. Support lets you help a specific public AEGIS workstream while also funding the core development work that keeps the project moving.
50%
Selected Fund
Reserved for the AEGIS workstream you name in the Ko-fi message, such as Search Bridge, Visual Training, Speed/Routing, Backend Safety, Benchmarks, or another public AEGIS target.
50%
Core AEGIS Development Fund
Supports public AEGIS development at the creator’s discretion, including tools, hardware, subscriptions, testing branches, maintenance, experiments, and future public releases.
Choose what to accelerate
Support what you want AEGIS to build faster.
Click through to Ko-fi. If you want to target support, write the workstream name in the message box. If no target is named, support is treated as Core AEGIS.
Roadmap
From careful chat to tested intelligence.
Funding target
AEGIS Search Bridge
Name this “Search Bridge” to support the hosted search layer: sharper Google queries, known-source lookups, and future AI-assisted rewrite capacity.
Leanfree/low-cost beta
Funding target
Image Generation
Name this “Image Generation” or “Visual Training” to support API-backed image prompts, edit specs, visual judgment, mismatch detection, identity/style boundaries, and critique calibration.
~$10kfirst major target
Funding target
Speed/Routing
Name this “Speed/Routing” to support compact runtime cards, effort-setting tests, context routing, and faster AEGIS behavior without losing discipline.
~$10klatency + cost work
Funding target
Backend Safety
Name this “Backend Safety” to support manual-mode boundaries, approval gates, action/result schemas, logs, replay tests, and safer paths toward future API automation.
~$10k–$15kruntime foundation
Funding target
Benchmarks
Name this “Benchmarks” to support testing across documents, PDFs, spreadsheets, plans, specs, critique, research framing, strict-output tasks, and reliable file delivery.
~$15k–$25kordinary-user lift
Research branch
Deployment/Embodiment
Name this “Deployment” or “Embodiment” to support controlled deployment research, simulated agent/device workflows, runtime safety patterns, and backend-dependent approval-gated operation.
Flexiblebackend-dependent branch
Experimental branch
Game-learning Research
Name this “Game-learning” to support optional creator-led research using games as repeatable environments for observation, rule discovery, context compression, uncertainty handling, and backend decision-making.
Flexiblecore fund / later support
Roadmap budgets are rough API/testing planning ranges. They are intended to fund evaluation and guide/runtime improvement. They are not promises of proprietary model-weight training or guaranteed capability milestones.