Philemon Kiprono 🇰🇪

AI Engineer | DSPy | Azure | ...Arsenal F.C

[Philemon Kiprono 🇰🇪 @ronoh4]

@NdindiNyoro Great view 🪟

[Omar Khattab @lateinteraction]

in case you missed it, OBLIQ-Bench is now on arXiv: https://t.co/RbL5aBOq4B

my hope is that this reduces the frequency of IR or search agents papers that I discard immediately as a reader because in 2026 they’re still evaluating on long-expired MS MARCO, NQ, HotPotQA, BEIR, etc

[Philemon Kiprono 🇰🇪 @ronoh4]

@scaling_shields Course

[Philemon Kiprono 🇰🇪 @ronoh4]

🚀 DSPy 3.2.0 is out!
🔗 BetterTogether chains optimizers: GEPA → BootstrapFinetune → GEPA via strategy strings
🔌 LiteLLM decoupling begins — custom backends, no litellm dep
🛡️ Hardened RLM & PythonInterpreter — structured errors, resilient parsing
Exciting times for #DSPy

[Philemon Kiprono 🇰🇪 @ronoh4]

@itsgabbyt @LagatJustin Excited to see you in Nairobi Kip Keino

[Charly Wargnier @DataChaz]

ANTHROPIC LITERALLY JUST HANDING US THE BLUEPRINT🤯

Their new 33-page guide on Claude Skills is the cheat code.

Make sure to bookmark this before it gets lost in your feed.

Link in 🧵↓

[Victor M @victormustar]

GLM-5.1 > Claude Code (Opus 4.6)?

I'm tripping or CC has become very bad but built a Three.js racing game to eval and it's extremely impressive. Thoughts:

- One-shot car physics with real drift mechanics (this is hard)

- My fav part: Awesome at self iterating (with no vision!) created 20+ Bun.WebView debugging tools to drive the car programmatically and read game state. Proved a winding bug with vector math without ever seeing the screen

- 531-line racing AI in a single write: 4 personalities, curvature map, racing lines, tactical drifting. Built telemetry tools to compare player vs AI speed curves and data-tuned parameters

- All assets from scratch: 3D models, procedural textures, sky shader, engine sounds, spatial AI audio!

- Can do hard math: proved road normals pointed DOWN via vector cross products, computed track curvature normalized by arc length to tune AI cornering speed

You are going to hear about this model a lot in the next months - open source let's go 🚀🚀

[elvis @omarsar0]

Pay attention to this one if you are building terminal-based coding agents.

OpenDev is an 81-page paper covering scaffolding, harness design, context engineering, and hard-won lessons from building CLI coding agents.

It introduces a compound AI system architecture with workload-specialized model routing, a dual-agent architecture separating planning from execution, lazy tool discovery, and adaptive context compaction.

The industry is shifting from IDE plugins to terminal-native agents.

Claude Code, Codex CLI, and others have proven the model works.

This paper formalizes the design patterns that make these systems reliable, covering topics like event-driven system reminders to counteract instruction fade-out, automated memory across sessions, and strict safety controls for autonomous operation.

Paper: https://t.co/tpAZFaSnog

Learn to build effective AI agents in our academy: https://t.co/1e8RZKs4uX

[Omar Khattab @lateinteraction]

.@LakshyAAAgrawal has done an unbelievably nice job collecting successful applications of GEPA to everything from state-of-the-art research to new products and established enterprises at this page.

Staying on top of these is harder than one would think!

https://t.co/2LTv6oFFrg

[Philemon Kiprono 🇰🇪 @ronoh4]

@Gatsz01 @Osama_otero Amazing

[Philemon Kiprono 🇰🇪 @ronoh4]

@tech_optimist Testin it out already

[Philemon Kiprono 🇰🇪 @ronoh4]

@lateinteraction Quite a feat 👏🏾👏🏾

[Philemon Kiprono 🇰🇪 @ronoh4]

@lateinteraction Sounds like a good weekend read

[Philemon Kiprono 🇰🇪 @ronoh4]

@lateinteraction I am testing it already and it has hidden features that are currently understated like how good it is in handling constraints enforcement for outputs

[Philemon Kiprono 🇰🇪 @ronoh4]

#RLM lets LMs run code to enforce constraints instead of “hoping” a prompt is followed. External #Python checks, logged & inspectable. More details here 👉https://t.co/FquMr4kD1Y

[Philemon Kiprono 🇰🇪 @ronoh4]

@a1zhang Amazing work @a1zhang I will be keenly following and contributing when time allows

[Philemon Kiprono 🇰🇪 @ronoh4]

An interesting take here. Is it possible we've been measuring intelligence invalidly?

[Robert Youssef @rryssf]

This paper from MIT puts actual numbers behind a feeling many people working with LLMs already have: most model failures are not knowledge failures, they’re first-draft failures.

The paper studies Recursive Language Models (RLMs) and asks a very specific question:

What happens if you let the same model revise its own output multiple times instead of scaling parameters?

The answer is surprisingly concrete.

Across reasoning-heavy benchmarks, the authors show that recursion consistently improves accuracy with no change in model size.

On multi-step reasoning tasks, adding just 2–4 recursive passes improves correctness by 10–25%, depending on task complexity. On longer planning problems, error rates drop even more sharply, because early logical mistakes get corrected in later passes instead of propagating forward.

One figure in the paper makes this especially clear.

They plot task accuracy vs. recursion depth.

The curve is steep at first:

• Pass 1 → baseline performance
• Pass 2 → large jump in correctness
• Pass 3–4 → diminishing but still meaningful gains

After ~4 iterations, returns taper off, which suggests something important: most reasoning failures happen early, and a small amount of structured revision fixes a large fraction of them.

There’s also a cost comparison that’s hard to ignore.

The authors compare:

• A larger non-recursive model
• A smaller recursive model using multiple passes

The recursive model reaches comparable or better accuracy while using fewer parameters and fewer total tokens in the final answer. Even though recursion adds compute internally, the output length shrinks because later passes compress and clean up earlier drafts.

In plain terms: the model thinks more, but talks less.

Another quantitative result I found fascinating is hallucination reduction.

The paper measures factual consistency across iterations and finds that later recursive passes explicitly remove unsupported claims introduced in earlier drafts. The probability of a hallucinated statement surviving to the final output drops significantly after the second pass, because the model is now evaluating its own content instead of blindly extending it.

This directly challenges the “long chain-of-thought = better reasoning” assumption.

The data suggests the opposite.

Better reasoning comes from iterative self-correction, not from dumping more intermediate tokens. The recursion acts like an internal verifier that gradually aligns the output with constraints imposed by the task.

There’s also a subtle systems insight hidden in the math.

If accuracy improves roughly logarithmically with recursion depth but model size improves accuracy sublinearly with parameter count, then recursion is simply a more efficient lever. You get more reasoning per unit of compute by looping than by scaling.

That’s a big deal for deployment.

Instead of asking:
“How big can we make the model?”

This paper asks:
“How many chances does the model get to be wrong before we trust it?”

The broader implication is hard to miss.

We’ve been benchmarking models on their first answer.

But intelligence doesn’t live in the first answer.
It lives in revision curves, error decay rates, and how quickly a system converges toward correctness when allowed to reflect.

This paper doesn’t just propose a technique.

It quietly suggests we’ve been measuring the wrong thing all along.

[Philemon Kiprono 🇰🇪 @ronoh4]

An interesting 'output' aspect of RLMs.

[Omar Khattab @lateinteraction]

Another understated aspect of RLMs: the *output* length is essentially unbounded too, not only input.

A simple test of the difference in expressive power between a Transformer and an RLM: Give your favorite model a 30k-token prompt and ask it to repeat it verbatim. They will all fail, but an RLM trivially succeeds.

[Philemon Kiprono 🇰🇪 @ronoh4]

@a1zhang @chaiyihein This is super cool

[Min Choi @minchoi]

This is literally my new workflow now:

Real-time search → Grok 4.1 Fast
Planning → Grok 4.1 Thinking
Frontend Coding → Gemini 3 Pro
Backend Coding → Claude Code (Opus/Sonnet 4.5)
Write Tests → Gemini 3 Pro
Run Tests → GPT-5.1 Codex
Debug → Claude Opus 4.5

Bookmark this.