[ ALL WRITING ]
Essays, articles, research talks & technical deep dives
Agent evaluation metrics: how to measure whether an agent works
A framework for choosing outcome, quality, cost, safety, behavior, and performance metrics that demonstrate whether an agent is actually completing its job.
p}]]W*>] ;jl(UQ_:jovMO'C\8zo(0Q_
IJ\a^nk'_mYvnpjjqh!oQox.8[Y|qipa
+{I?mW~czO<c@#_@J]xB>_vI"n`"M?@~
X(-d YYjU[\jwof\>@BcY`pa|(vwoo''
^_0wn)|jvlxZ,x ?^W1UQ~M|qZ-[aatJ
M*!U~l>U`)]_)^(;&^/tZl_p}:qW1/1~
+;ah@a'||lB{wU]]{W M^w@B`-O^ L/^
[fa|j}%m<qpkJ$]QM|CC/[&%o`Mr0)\0 BETA DISTRIBUTION
Agent observability: how to trace, debug, and improve AI agents
A practical guide to capturing agent traces, connecting execution across sessions, and using evaluations to diagnose and improve production AI agents.
{${${${${
{ { { {{ {$Q$Q$Q$Q${{
{ {${${ { {${{$${$QQQQQQQQQ$${
{${$Q$Q${${{$Q$$QQ$QQQQQQQQQQQQ$
$Q$QQQQQ$Q$$QQQQQQQQQQQQQQQQQQQQ
QQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQ
QQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQQ
{{{{{{{{{{{{{{{{{{{{{{{{{{{{{{{{ MARKOV CHAIN
How to evaluate AI agents: a production workflow
A production workflow for evaluating agent outcomes, trajectories, individual decisions, and reliability, then turning failures into regression tests.
>>>>>>> > > > >>>> >
> >> > > > >> > > > >
> > > > >>> >>
Y > > > > > > > >|
$ b > >> >> | > > |>>>> |
| Y | > >>>Y>|>|>|> >||||> Y
Y Y >|>| Y |> >|>>>| >>>>>>>
> || > > ||| >> BROWNIAN MOTION
AI agent frameworks compared: LangGraph, CrewAI, AutoGen, and more
A practical comparison of agent SDKs, orchestration runtimes, harnesses, and managed platforms for production AI systems.
>> > > > >> | > >>> >> >> > | | Y > | > | > > >> >>> > b Y Y Y | > Y > > > | >> >>> $ | > | > | Y>|>Y > > > | |> > | > >>>Y>>>|>| > > | > > | > | >>> >>>> >> > | > > > ||> >>> >>>> >>>>>
BROWNIAN MOTION
Best AI Observability Tools for Autonomous Agents in 2026
A comparative look at leading observability tools for autonomous AI agents, with evaluation criteria for production deployment in 2026.
?i!i">>i+_+{+!-[+_+':+"><!,`.><!
i;l>l,>li!l>I]!I:>~>iI<`:>>+<:i`
.->-,;^:!_}[f|)\xX|(~1)<}<,><"~!
?,+-l??+}\njzOCJLqpUcf|[1~[_<i I
!_><{!}/Xruj#*aB$kZZ0zUtv/[?}i><
l?>]-?<]}jr|QLzqY0Czr|Y1?[?!>!?<
^_I~~+1:]+1]rjr|ff~\<[~|+-<>~i+?
> i>>{,il[~<!+-+?<1?<l+]~:~l><,l GAUSSIAN DISTRIBUTION
Understanding Latency in AI Model Deployment
A practical breakdown of where latency comes from in deployed AI systems and how teams can reduce bottlenecks across the inference path.
^ ' '.`' '''`''` .'`'. '`'` ''
-+ii;,;^`"^`.^^ .^' `^ .`'`` '^
v/[]>!;l;"`'^"^'`^.`^`.'.` `'.'`
bCn\]-<>:;"""`^^`..``.. `'^.^'^'
$kUx\[?_li;"^,`'```'^^' ^.. . .
wCr\]]+<Il:;`^'''^'.`' .'`` ^
n|{+~i;I";^`"`^`^..'`' ` ` `
_+>I:I;^""^.`` .. .^.``'^`^ .^ EXPONENTIAL DECAY
Rented AI and Hidden Costs
An analysis of the hidden operational and business costs teams face when relying on rented AI stacks without strong infrastructure ownership.
'. '`' ''^'.'.`^ ^^ '`'. ''.^
-~~!I;,","'^^` ^''^'` .' `' '.'
c/{]<~!I;,^``''.^ '^`'`..'^ ` .
pUn/}++!!lI"````^.^^^..''`.`^ '^
$aQx\}_~>II,":"^''''. '^ ^''.''.
dYr/[-~i!:,":'^`'""^^ `.''. '^.'
cf{_~ilI,;,"'`^.^` ^'` ''`. `..'
]<!!l"^::'^^`'`'.^' ^`.`...'`'`. EXPONENTIAL DECAY
Why AI Agents Break: A Field Analysis of Production Failures
A comprehensive field analysis of production failures in AI agents. Explores the root causes and failure patterns that occur when AI agents break in real-world deployments.
`..`. ''.^^^::;II;I;^"'`...''`'
'`.. ''^`.",I_~]1}}]~il,"`'' ``.
. .''``.""l;+/\cUYcct]-I`".'`.'
'' '.'.^,,i<1XXb*adpX()>,^^^.` '
.. '``;I~<\OLM$%##0/|~;I`^`''.
'.'. '`",>i|UYd*odbY((i;;^'.'.
'...'.`"ll+||vYYuu\?-!^``'`'`.
. ` .^^'":!_~]{[?}+ll"`^.` `.. BINOMIAL DISTRIBUTION
Observability-Driven Sandboxing
Securing AI agents through observability-driven sandboxing: techniques for monitoring and controlling agent behavior.
`'.`. ``^''.' ..`'``'``' '. `'..
?_<i:",`,`"'. . . '. .''`'. `.
v([]~>;:;:^`^'"." '` `` `. ''. ^
mXx(]?<!!I,"^,.^`^'^` .`.^ `.'
$kJr\1_<<!lI^"^^.'..`^ ^`` `.^'.
mUx1}_<ll:I:"^"".`.`^`. .`^..`^
v({]<l!I;;^''`'''`^` ` `^'' . .`
]<iI,",`^'^..^.^.`.^. `'^. ..`.` EXPONENTIAL DECAY
Agent Frameworks
An overview of popular agent frameworks and multi-agent frameworks. Explores how AI agent frameworks act as control layers around models, managing execution order, tool invocation, state changes, and component interactions.
~+^Zx_!}cwm<BB#xhdoo(]~- .bbdLu1
ok!(jm>Zk8Qxq `!$L+:z?\-[}? O Xp
~,-|qmJax\Uw&-u]b}x"!tInaj]mn-dX
cZI*):cqZimkrk|+-!@|&C\ztd0|M--M
d<: ,n}1 _fwiqQ0h[*" (r|J>/^do\J
YUWYk0*!u@;~ti>|{:[WXiiq{>b; Z\;
[oU&r|@({O@zX+?rj"(nJ{fl 8q/"x@&
!'_) Y_~#o@Y;lX- <8(}i;vM1iU>cZz BETA DISTRIBUTION
Top 5 AI Prompt Management Tools of 2025
Compare the best prompt management platforms — Arize AX, Arize Phoenix, PromptLayer, Langfuse, and PromptHub — to find the right fit for your team's workflow and scale.
wh\t";.hbrv*powzJf>IudL-[x8],dm;
_{{z(z]wv,u?><o<uMQh?bpp<"%J+mXc
/QC)c vOIo+]x8;%i_nbWI|[XL:U"YX
bp.++Z_xr;ru)pY[>kljm$bd`<aZl,dz
jJO[{WMqrm}j(xL[_XZ8q_][,[/{}t0~
Or"a)bk(;xr~:?{a~&)c/{[!#-!d;)lt
`(b1i q_Z)'v:-CLWjX\)Yx>Z>Q;^::+
;+zr CU:|U>!/-_~?pkI@QfUQ(?tntYl BETA DISTRIBUTION
Benchmarks Are Broken: 5 Operator Metrics
Explore the 5 critical metrics AI executives should track beyond traditional benchmarks to measure real operational performance.
^`.^^.'.^'^.`'` . ''.^^ ^.''^' [+!!l,^^"^'"'"..'``^` ^.`' '^'.` u/)]<!;,I:":"^.^^` ``^.' `'` dLx|1_>iI,I:"".'".^.. `.`^`'. ^ $a0n/[_+!!l:;^","'^.'''.`. ^^'`' dCj/}]>>;I^,^,^`"`'` ` .``.^``^` x|1[+!ll,:`^'`..'' '`^ ``'.`^ '' _<>iI"^"""`^``^`''^. . ` '.`^`
EXPONENTIAL DECAY
The Platform Gravity Problem
Analyzing why the middle market is collapsing in AI and how platform gravity is reshaping competitive dynamics.
xCwq0c|?+?|c0qwCx1-_}fYZqZYf}_-1
]!;i{cpWB#Zr-;"l[rLqpOYx//fuzXvt
w*MaQt<` "-nwM%#wv)_~?)jvcvrffxY
/?<~}rQbadCj]>>?/XZqZUr)[{/vC0Lz
1nCOOJut|\rXLQUx1+>+(cw*MkC/+;:<
qYt{]}|jnxf\(\rUwaaqX)!. ;{XkB$M
1nCOOJut|\rXLQUx1+>+(cw*MkC/+;:<
/?<~}rQbadCj]>>?/XZqZUr)[{/vC0Lz WAVE INTERFERENCE
Frugal Innovation in AI Products
How constraints and frugal innovation principles lead to better, more efficient AI products in competitive markets.
jvJ0qbao*oabq0Jvj([+iI,^^^,Ii+[(
xt({]_~<>><+-]1|fxcYCQOZmZO0CYcn
\\\\\///ttfjjrrxnnnuuuuuuuuunnxx
cUQZqpbbbdqZLUvr\1?~iI:,,,;l>_[)
{->;"' '^;>-1fvJZb*&%@$@%Wap0Y
JQmpbhahhbqZLYuf([+i;"`'''^,l<-{
??]][}{)(\/frnuczYUJCCLLCCJJUXzv
0Jcxt|1[?_++_-?[{(\tjxuvvvvunrf/ WAVE INTERFERENCE
Top 7 Agent Supervision Platforms
Comparative analysis of the best AI agent supervision platforms available in 2025 for production systems, featuring 15+ min read deep dive.
. .
..^^Il++}}((((({{-_>>:;""`'..
.. '`::_+//YYZZOOOUUrr{{<<;I^"''
''II}{czddWWWMWbkJUtf??!!,,^`
. ``ll11YYaa@@@@$**QQrj][ii::`^
`'lI[[zcddWWWWMbbJJtt??!!:,^`
. '`::__//YUOOOOOUUrr1{<~;I""'`
. '.^^II++{}((((|{{-_i>;;""`'.. POISSON DISTRIBUTION
DeepSeek-R1: Redefining RL in AI
Deep dive into how DeepSeek-R1 is revolutionizing reinforcement learning approaches in modern AI systems.
{ {
{{${${
$$Q$Q${ {{
QQQQQQ${{{ { { {{ {{$$
QQQQQQQ$$${{${{{{{${{{{{{$${$$QQ
{{{{{{{{{{QQ{QQQQQ{QQQQQQ{{Q{{{{ MARKOV CHAIN
How to Build a RAG Chatbot
Step-by-step engineering guide to building production-ready RAG-powered chatbot applications.
nzUQOwqpdpqwOQUznj\)}]-___-]})\j
qqwmZOQLJYzuxjt\(1}[]????]][{1(\
]_<!l;;Ili~?{\rzCZdaM8B@$@%W*bwL
jjrrrrrxxxnnuuuvvvccccccccccvvvv
pkoMW&&WMokq0Uuf)]<l,`. '"I>-1
}}{{1)(\/frnucXYJCQ00OOOOO0QLCJY
t({]_+<<<~_]{(txzJ0mdkaoooabpZQY
dppqwZOQCYzuxf/|){[]--___-?][{)| WAVE INTERFERENCE