VSvarunsingla.com

← All entries

Day 106· · 3 min read

100 Days of Agentic AI -- The Milestone

Foundations & Protocols

protocols, memory, identity, evals, economics, twenty-two industry verticals and a frontier that reorganised itself along the way. Today we grade our own calls -- what the series got right and what it under-called -- distil the ten technologies that mattered most, and set the curriculum for the next hundred days.

Viral app of the day

caveman by JuliusBrussee

84K+ GitHub stars, #2 GitHub Trending; one-file SKILL.md for Claude Code/Codex/Gemini/Cursor + 30 more tools that cuts ~65-75% output tokens by making the agent talk like a caveman -- same answers, never touches code; used by devs at Nvidia + GitHub; perfect Day-100 punchline after tokenizer inflation and tokenmaxxing: 'why use many token when few token do trick'

By the numbers
100
issues in 108 calendar days
97M
MCP monthly SDK downloads (~970x in 18 mo)
~37%
lab-to-prod performance gap (real vs benchmark)
~1,000x
inference cost collapse over 3 years

1. Grading the calls: five theses under the microscope

What we under-called: (1) how fast frontier releases became government-mediated -- Fable 5's 19-day suspension and export-control-gated restore, GPT-5.6's ~20-org government-approved preview and the White House voluntary release standards all arrived faster than any issue predicted; (2) our own drift -- Days 64-98 ran a vertical-a-day until the standing rule pulled the series back to technology-first at Day 99. Both logged; the second correction was Varun's call, and it was right.

The callVerdictThe receipt
Compression thesis (Day 80): value moves to distribution + skills + governance, not benchmarksCONFIRMEDSonnet 5 matches Opus 4.8 at $2/$10 promo; new tokenizer (+~30% tokens) makes price sheets incomparable; procurement opens with 'show me your audit trail'
Aug 2 two clocks (Day 81): build to the original date while the Omnibus stays unadoptedVINDICATEDAt T-27 the deferral is still not in the Official Journal; Article 50 + governance machinery + GPAI penalties bind Aug 2 regardless
IPO race (Day 63+): frontier labs go public on unit economicsCONFIRMEDAnthropic S-1 Jun 1 (~$965B, Oct Nasdaq); OpenAI Jun 8 (weighing 2027); SpaceX's ~$75B raise = template; Anthropic overtook OpenAI on revenue (~$47B run-rate)
Protocol stack (Day 86): MCP/A2A/SKILL.md become the railsWONMCP ~97M monthly downloads, 59K+ servers; A2A in 150+ production orgs; SKILL.md in 16+ tools; agent-callable is the new indexed
Vertical playbook (Days 64-98): two loops, human on the commit seam, one audit trailGENERALISEDThe same pattern held in all 22 verticals; governance-by-construction exemplars (Mindtrip, Zillow, Bedrock Operator, Carbon ATK) won their categories

2. The ten technologies that mattered most

#TechnologyWhy it mattered
1MCP + the protocol stack (A2A, AG-UI, x402)The connective tissue that turned isolated agents into an economy; ~97M monthly downloads, 59K+ servers
2Agent Skills / SKILL.mdThe unit of agentic value and the new app store -- one artefact, four channels; allowed-tools = the trust boundary
3Long-horizon agentic RLPlanner/executor factorisation + process-and-outcome rewards; the technique under Sonnet 5, GPT-5.6 Sol and Qwen's 35-hour, 1,158-call run
4The agent memory stackWrite-Aside, MemOS, MemRL, transactive memory -- the skill set survives model rotation, so memory (not the model) is the moat
5Agent identity: SPIFFE/SVID + KYAWorkload identity, scoped authority, kill switch as identity protocol -- a fleet you run vs a fleet that runs you
6The streaming data planeMaterialised views as working memory, fresh by construction; ~90% of new TiDB clusters now created by agents
7Eval 2.0: trajectory + self-healing loopsScore the trajectory, not just the output; closed golden-set loops = the answer to the ~37% lab-to-prod gap
8Inference economicsFP8, speculative decoding, vLLM-class serving, custom silicon: the ~1,000x token-cost collapse, governed by cost-per-successful-task
9Capability gating as release architectureSame weights, classifier-gated safeguards, usage credits, revocable deployment -- Fable/Mythos wrote the frontier's release template
10Physical AI: VLA + actuation envelopesGeofence, rate caps, remote takeover, hardware E-stop -- agents off the browser and onto excavators, tractors and factory floors

3. What 100 days taught about how agents actually ship

Two loops at two speeds. Across all twenty-two verticals the same pattern held: a high-volume document or advisory loop automates first -- observable, reversible, measurable -- while a high-stakes commit loop keeps a human on the seam: the diagnosis, the price, the purchase order, the master, the SAR filing, the live-network change. Every successful deployment drew that line explicitly; every governance incident of the last hundred

One trail, three audiences. The OTEL gen_ai spans streamed into a WORM store for the regulator (Annex III evidence) turned out to be the same artefact procurement demands in the RFP (Day 82) and the same telemetry FinOps prices from (Day 83). Wire it once, reuse it three times -- the cheapest compliance decision of 2026 and the most repeated recommendation in this series. Reliability is where the value lives. A ~37% benchmark-to-production gap; 51% claiming production while only ~10% run true multi-agent systems. The boring disciplines -- SLOs, error budgets, circuit breakers, golden sets, chaos drills -- separated line items from demos. A governed agent you can audit beats a smarter agent you can't. One hundred issues later,

#TechnologyWhy it mattered
1MCP + the protocol stack (A2A, AG-UI, x402)The connective tissue that turned isolated agents into an economy; ~97M monthly downloads, 59K+ servers
2Agent Skills / SKILL.mdThe unit of agentic value and the new app store -- one artefact, four channels; allowed-tools = the trust boundary
3Long-horizon agentic RLPlanner/executor factorisation + process-and-outcome rewards; the technique under Sonnet 5, GPT-5.6 Sol and Qwen's 35-hour, 1,158-call run
4The agent memory stackWrite-Aside, MemOS, MemRL, transactive memory -- the skill set survives model rotation, so memory (not the model) is the moat
5Agent identity: SPIFFE/SVID + KYAWorkload identity, scoped authority, kill switch as identity protocol -- a fleet you run vs a fleet that runs you
6The streaming data planeMaterialised views as working memory, fresh by construction; ~90% of new TiDB clusters now created by agents
7Eval 2.0: trajectory + self-healing loopsScore the trajectory, not just the output; closed golden-set loops = the answer to the ~37% lab-to-prod gap
8Inference economicsFP8, speculative decoding, vLLM-class serving, custom silicon: the ~1,000x token-cost collapse, governed by cost-per-successful-task
9Capability gating as release architectureSame weights, classifier-gated safeguards, usage credits, revocable deployment -- Fable/Mythos wrote the frontier's release template
10Physical AI: VLA + actuation envelopesGeofence, rate caps, remote takeover, hardware E-stop -- agents off the browser and onto excavators, tractors and factory floors

5. The next 100 days: the curriculum

The pipeline, technology-first per the standing rule: AI-for-science architectures (Claude Science and lab-in-the-loop agents -- promoted to next issue by this week's launch); self-improving agents 2.0 (MemRL in production, skill-library evolution, Darwin-Godel updates); world models and simulation-first agents; MCP 2.x / A2A 1.x and the interop spec; context engineering at native 1M tokens; neuro-symbolic and 100x energy-efficient inference; on-device agent hardware and agent-first phones; frontier-safety tech; and the next

The watchlist that decides H2: August 2 enforcement (T-27) and whether the Omnibus reaches the Official Journal in time; Anthropic's October Nasdaq debut and whether OpenAI holds 2026; the first Skills-marketplace supply-chain incident (this series' standing call for H2); GPT-5.6 reaching GA beyond the government preview; and whether the White House voluntary standards harden into rules. The method for the next hundred stays the same: learn the technology under the headline, wire the governance before the scale, and keep score in public.

Market signal: one hundred issues in, the scoreboard reads: models compressed (~3%), tokens collapsed

(~1,000x), rails standardised (MCP/A2A/SKILL.md), and the frontier became government-mediated. The durable moats are exactly the ones the series kept finding -- distribution, proprietary data flywheels, governance evidence, and the lowest cost per successful, audited task. The next hundred days are about who operationalises that fastest: in public markets, under enforcement, at scale.

Viral App of the Week: caveman (by JuliusBrussee)

The token-diet skill that became a movement: a one-file skill/plugin for Claude Code, Codex CLI, Gemini, Cursor and 30+ other agents that instructs the model to 'talk like caveman' -- drop filler, keep substance, use fragments, never touch code, commands or errors. The same technical answers at ~65-75% fewer output tokens, in use by developers at Nvidia and GitHub. The perfect Day-100 punchline: after a year of tokenizer inflation, usage credits and tokenmaxxing budget burns, the community's hottest agent upgrade is teaching it to say less. Tagline: 'why use many token when few token do trick.'

Market signal

one hundred issues in, the scoreboard reads: models compressed (~3%), tokens collapsed (~1,000x), rails standardised (MCP/A2A/SKILL.md), and the frontier became government-mediated. The durable moats are exactly the ones the series kept finding -- distribution, proprietary data flywheels, governance evidence, and the lowest cost per successful, audited task. The next hundred days are about who operationalises that fastest: in public markets, under enforcement, at scale.

Practical takeaways
Tomorrow -- Day 101: Claude Science and the lab-in-the-loop

Tomorrow -- Day 101: Claude Science and the lab-in-the-loop stack: how AI-for-science architectures actually work -- agentic literature synthesis, hypothesis generation, self-driving-lab integration, and what a drug-discovery agent for neglected diseases does under the hood.

VS
Varun Singla
Singapore · About · Learning in public