OpenAI ships extensions for ChatGPT - that feed more to plug into the consumer experience. Anthropic ships extensions (made for builders) into ClaudeCode - feel more DX.
226 karma · joined March 22, 2020
OpenAI ships extensions for ChatGPT - that feed more to plug into the consumer experience. Anthropic ships extensions (made for builders) into ClaudeCode - feel more DX.
Feels like fair bit of overlap here. It's ok to proceed in a direction where you are upgrading the spec and enabling claude wth additional capabilities. But one can pretty much use any of these approaches and end up with the same capability for an agent.
Right now feels like a ux upgrade from mcp where you need a json but instead can use a markdown in a file / folder and provide multi-modal inputs.
It also opens up - ads - subscriptions - transactions
based revenue lines for OpenAI
- Need better memory management and controls (especially across multi-repos) - /upgrade needs better management
1. Why do you want to fragment context? Why not add this to the main file with inline docs?
I also wonder whether software products can be discovered through a launch any longer. Just too much noise and high throughput at which they are being released - once u get past the page all products look like they were baked by clones or a completely divergent one that needs investment.
The launch model is broken.
1. Break down the feature or bug report into a technical implementation spec. Add in COT for the splits. 2. Verify the implementation spec. Feed reviews back to your original agent that has created the spec. Edit, merge, integrate feedback. 3. Transform implementation spec into an implementation plan - logically split into modules look at dependency chain. 4. Build, test and integrate continuously with coding agents 5. Squash the commits if needed into a single one for the whole feature.
Generally has worked well as a process when working on a complex feature. You can add in HITL at each stage if you need more verification.
For larger codebases always maintain an ARCHITECTURE.md and for larger modules a DESIGN.md
I see 2 big factors to improve ability of coding agents today
- on device model - context (or understanding of modules) - not only retrieving the relevant sections or codebase but creating a version (transforming it) which is readily consumable by a model and used to focus on the problem at hand.
This requires both a macro global context of the codebase and the ability to retrieve the local context of the problem being solved.
Augment context e.g. does a fairly good job of context compression and retrieval among coding agents. Fast indexing & retrieval is a good step forward to enable open context compression
All products in this space by YC teams are targeted at scaled voice agent startups or teams.
- Roark (https://roark.ai/)
- Hammin (https://hamming.ai/)
- Coval (https://www.coval.dev/)
- Vocera (https://www.vocera.ai/)
How do you differentiate - who is this for? Voice agent devs paying $500/mo. for early stage software?
What do you think of 1. Licenses 2. Air traffic control & routing for personal crafts 3. Landing / Takeoff / Turbulence (weather) - where typically most of the human skill today is used in for commercial flights. 4. and probably the one for interface design and overall autonomous systems - human decision making + autonomous systems really have a difficult time working in tandem - fully autonomous is less risky at an aggregate level but puts onus and more of the blame on the company / maker versus human-in-the-loop.
I would love to see this article also expand to touch upon things like : - data management - (tooling, frameworks, open vs closed data management, labelling & annotations) - inference as a pipeline - frameworks for breaking down model inference into smaller tasks & combining outputs (do DAG's have a role to play here?) - prompts - areas like caching, management, versioning, evaluations - model observability - tokens, costs, latency, drift? - evals for multimodality - how do we tackle evals here which in turn can go into loops e.g. quality of audio, speech or visual outputs
Perfect recall, no need for memory, available at your fingertips - with T&C that completely disregards your need for personal privacy.
The challenge though is the "better" alternatives where a mix of privacy and convenience is there is not always convenient. We have and continue to be marketed convenience at the cost of privacy.
What we need is consumer data privacy to really become a societal and government concern and for devices which infringe on this to really go through the same scruity as say a drug going thru FDA approval.
Would love to see if this is the same when reading different patterns of code - functional vs object oriented. Where do product specs or technical specs lie?
All of this feels more pseudo-science to me - the distinction between language and math is not well understood at the moment.
More often than not critical parameters or mis-configurations happen because of internal tools which work on unpublished params.
Internal tools should be treated as tech debt. You won't be able to eliminate issues but vastly reduce the surface area of errors.
Tools will no longer need operators as adding multi modality to tools will ensure we can actually have the operator interface on top of the tool.
Tuning the operator is where the playing field shifts to as operators which can translate fuzzy inputs to creative specs or workflows which can in turn be executed by the downstream engine.
Hitherto we have always thought of software as one for many - the aspects of adoption of software is left to humans.
Maybe we will shift more of our skills to higher level skills - how do we make more enjoyable products rather than how do we learn the nuance of programming languages, software paradigms and we may evolve software and technical paradigms around these.
I doubt this is the age of the death of building. We will just invent and build new things.
Generative Media by AI - while dystopian can also make us think what does media really means - will there be more value on in-person media rather than digital media?
Will it also mean we have a larger spectrum of artists who can now create - not just books and blogs but multi-media experiences that are immersive worlds/moments?
While we can debate the merits/demerits of push - for practical commerce, a large part of the world still switches on push, and relevant push will deliver more concrete and better ROI.