Favur Dev (@favurdev) header
Favur Dev (@favurdev) profile picture

We run multi-agent teams that take a SOW and ship a repo. Public failure notes from the runs. Evals every Friday. Invite-only.

United States
Joined June 2026
Photo by Favur Dev (@favurdev)Photo by Favur Dev (@favurdev)Photo by Favur Dev (@favurdev)Photo by Favur Dev (@favurdev)Photo by Favur Dev (@favurdev)Photo by Favur Dev (@favurdev)Photo by Favur Dev (@favurdev)Photo by Favur Dev (@favurdev)Photo by Favur Dev (@favurdev)Photo by Favur Dev (@favurdev)
In honor of the leather jacket. Land on your feet Jensen 🫡 What happened at NVIDIA and the future of AI and chips
1
2
4
18,197
We tested idiotproof with a client video we made a few days ago. Very nice, @wordgrammer! Now Opus can drive the workflow end to end, with idiotproof handling the rendering and no extra Fable tokens.
1
2
4
388
Favur Dev retweeted
Replying to @iscream3d
Full repo here: github.com/FavurDev/WasmRend… Requests: 41,081 Input Tok: 3,320,399,117 Output Tok: 20,021,990 Reasoning Tok: 6,691,274 Cost: $756 Models: @Zai_org GLM 5.3 @AIatMeta Muse Spark 1.3 @GoogleDeepMind Gemini 3.8 Flash
1
1
2
26
Introducing Gemini 4 Argon – our new frontier model. It’s built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense – rolling out today to a set of trusted testers through our Fairwind Program.
1,980
4,609
4,525
42,688
9,062,091
Favur Dev retweeted
Most fixes for context compaction try to make the eviction judgment sharper. A pruning tool for the model, a smaller model scoring relevance, layered summaries. All of them run that judgment at the one moment the system knows least. By then it is four hundred messages of undifferentiated text, and length and recency are the only properties left, both anti-correlated with what matters. Favur is our team of agents that builds software from a written specification, and in it every message gets a retention class at append time, chosen by the call site that knows why the message exists. Forgettable, fuzzy, strict. Compaction reads the classification, never the content, and summarizes a class one tier before it drops it. The trade is that a compaction routine is wrong in one findable place, and retention classes are wrong at whichever call site forgot. That failure is silent by construction. The tiers, the thresholds, and what the design costs: dev.to/favur/deciding-what-a…
6
5
93
673,928
1st Place: Nova taught Grok to reverse-engineer binaries into clean C. Started with 262KB GameBoy ROMs, ended up rebuilding a 1995 car’s ECU runtime. @theoc____ @supratikp07 @henryzhang
110
292
33
4,690
605,473
Favur Dev retweeted
We shipped a major upgrade to the Auto Router: openrouter.ai/openrouter/aut… It now routes based on what the market actually uses for each type of task. When the market migrates a workload to a new model, the router follows within days. Benchmarks & more 👇
33
21
6
292
40,816
We score how each AI model performs across a full software lifecycle — and no one model wins it all. Different models led on quality, cost, and speed. The benchmark behind @favurdev — scored from run artifacts, not by an AI. What we found: dev.to/favur/which-ai-model-…
1
1
1
73
Favur Dev retweeted
The prompt engineering here is super impressive! Such a great example of agent prompting:
GPT-5.6 Sol Ultra produced a proof of a 50 year old math conjecture. Unlike the Erdős Unit Distance Problem, this was done with a model publicly available *today*. I look forward to seeing what scientists and researchers are able to do with this model!
23
76
14
1,463
145,019
we build custom software for small businesses — internal tools, automations, the odd full MVP, plain descriptions, full source code every time. also the makers of Favur, our multi-agent dev team, more soon. first time saying it publicly: awesoftsolutions.com/?xcard=…
4
1
3
153